I got into government reform sixteen years ago, though I didn’t think of it as reform at the time. I thought of it as just trying to make a few specific things work better. Since then I’ve worked at the local, state, and federal levels, on benefit delivery, on national defense, on a handful of things in between. But the government we have — the operating model it runs on, the rules and structures and assumptions that shape how it hires, procures, and delivers — was built for a world that no longer exists, and the distance between that world and this one is growing. We are approaching the kind of moment when that gap stops being a management problem and becomes a true legitimacy crisis. It’s time to start asking whether the theory of change most of us have been operating under — incremental improvements off a pretty poor baseline — was ever going to get us to a government capable of meeting fast-changing needs. It hasn’t yet, and if we don’t do something differently, it won’t.

Subscribe now

Kelly Born at the Packard Foundation recently shared with me a framework called the Three Horizons, originally developed by Anthony Hodgson and adapted widely in systems-change work. In it, Horizon 1 is the currently dominant system. It’s functional enough to persist but failing in critical ways, especially for people with less power.1 Horizon 3 is the future system you’re working toward, already visible in patches of practice that embody different values and different ways of working, but far from the norm.2 Horizon 2 is the turbulent middle where change agents work.

But the key insight is that not all Horizon 2 work is the same. Some H2 innovations genuinely create the conditions for the new system to emerge. Call those transforming H2, or H2+. Others, however inadvertently, extend the lifespan of the failing system by relieving the pressure that might otherwise force structural change. Call those sustaining H2, or H2-. Both feel like reform, but they have very different long-term implications.

A lot of what the government reform field has built over the past fifteen years, including a lot of my own work, is probably H2-. Not because the people doing it were wrong, and not because the work didn’t matter, but because the structure of the work, no matter how well executed, had a systematic tendency to solve the immediate problem in a way that made the structural problem easier to tolerate. If that sounds like an indictment, the first place I’d point the finger is at myself. And I don’t regret any of it. Nor am I saying that there’s not a valuable place for H2- work. There is. I’d just like us to see it for what it has been so we can be more deliberate about what we do next.

There are a few patterns that play out over and over in the category of H2-, the work that sustains the status quo.

The pressure valve

The dynamic of unintentionally propping up the status quo is easiest to see in what you might call vertical interventions. An agency lacks the talent to do important work, so philanthropy pays for some detailees, often through an Intergovernmental Personnel Agreement. An agency can’t hire the people it needs under standard civil service rules, so Congress grants a particular team a special hiring authority, as it did for the CHIPS implementation team. The procurement rules will take too long, so an agency gets an Other Transaction Authority exemption. The answer to “the system doesn’t work” has been to hack it in the vertical domains that get attention, not fix it horizontally.

Each of these carve-outs delivers genuine value. Most of them are the right call in the moment. But each one also functions as a pressure valve on a system that, absent the relief, might generate the political and institutional pressure needed to fix the underlying problem. The legislative carveout that makes Paperwork Reduction Act compliance unnecessary for a particular program means the people and organizations with the most standing to demand PRA reform never do. They just move on to the next challenge. The rescue team that fixes a broken system well enough to make it functional returns it to “good enough,” which is exactly the condition in which structural change is hardest to achieve.

The cost of the workaround isn't just lost pressure. Every special authority, every exemption, every vertical fix makes the overall system more complex, more fragmented, and more navigable by sophisticated actors — large contractors, well-resourced agencies, organizations with the staff to learn which door to knock on. The organizations least able to navigate that complexity are typically the ones serving the populations with the worst outcomes. Vertical interventions don't just leave the underlying dysfunction in place. They tend to entrench it.3

The demonstration that doesn’t scale

A close cousin is the successful pilot, a project that shows that SNAP enrollment can be simplified, agencies can build digital services in-house, or test-and-learn mechanisms can improve program outcomes. The pilot works. But because the structural conditions for scaling don't exist, it stays a demonstration, or it gets scaled in a form that loses the properties that made it work, because the procurement rules, the civil service rules, or the oversight structures were designed for a different kind of work.

These pockets of better practice are real Horizon 3 evidence. They become H2- when the political system treats them as proof that the problem is solved rather than proof that the problem is structural. Demonstrations that show something is possible but don’t change the conditions that make it the exception (especially in the four competencies I’ve talked about before) can end up providing the illusion of progress.4 Philanthropy reinforces this: demonstrations are fundable because they produce legible, attributable, near-term outcomes. Horizontal structural reform — changing civil service rules, repealing the PRA, restructuring how digital projects get budgeted — produces outcomes that are diffuse, slow, and hard to attribute. The result is a field well-resourced to demonstrate what’s possible and under-resourced to change the conditions that keep the possible from becoming normal.

The co-optation trap

The third pattern may be the most pernicious. The field promotes concepts that carry genuine Horizon 3 values — user-centered design, agile development, evidence-based policy. And then the H1 system absorbs the vocabulary, where it does the work of legitimizing the status quo rather than challenging it. "Agile" is now a procurement category that large contractors have learned to perform while delivering waterfall outcomes on agile timelines. "Human-centered design" is a workshop format and a contract deliverable. When evidence production is mandated as a compliance activity, it can crowd out evidence as a genuine decision-making tool. An agency that goes looking for the best possible data because it wants to know something is in a very different relationship with that data than an agency that produces a required report because OMB or Congress needed the box checked. From the outside the two can look almost identical. They produce very different organizational behavior.

The deepest version of co-optation is when the system doesn't just absorb the vocabulary of reform, it learns to perform it. An agency can run user research, adopt agile ceremonies, and publish evidence reports while leaving the governance structures, vendor relationships, and incentive systems that produce bad outcomes entirely intact. The performance is often sincere. The people doing it believe in what the words mean. But copying the visible practices of a different way of working, without changing who has power, how decisions get made, and what anyone is accountable for, produces a somewhat nicer version of the same thing rather than a different thing. Reform becomes a ritual that signals values but doesn't deliver on them. The system learns to speak the language, but it doesn't actually change.

Share

A different moment

The H2- work I’m describing has been done in good faith by people. I am one of those people. Code for America, which I founded and where I spent more than a decade, is in important respects capacity substitution. For much of the past fifteen years, the H2- path was arguably the right call. When there was no political space for structural change, demonstrations were a good way to build the evidence base and develop the field.

I think we are in a different moment now. This moment is defined by disruption. I count three kinds:

Contingent disruption — pandemics, climate events, geopolitical shocks, financial crises — is unpredictable in its specifics but very predictable in its category: large, fast-moving, high-stakes demands that fall disproportionately on government. COVID was not an anomaly. The next version won’t look the same.

The most recent disruption to federal government was political. Whatever the cost of its methods, DOGE made the brittleness of the current operating model impossible to ignore and created political openings for structural arguments that previously had no traction. The reform field did not create this moment. But it can shape what comes out of it.

AI brings structural disruption — a transformation already underway in the material conditions of work, economy, and administration. AI is not only an exogenous shock that government will have to absorb. It is also moving the bar on what counts as acceptable service in the first place. People are already using AI to understand their medical bills, navigate insurance denials, and draft appeals for benefits they were wrongly denied. Soon they will expect to apply for SNAP or file their taxes by uploading a paystub and answering a few plain-language questions. The forty-page PDF used to feel intolerable. The well-designed web form will start to feel that way too, and faster than the last transition did. And service delivery is only the most visible piece. If a small team with the right tools can map a regulatory regime in a week, the timelines we have now, in which rulemaking takes several years — or even multiple presidential terms — become indefensible.

So the gap we have been measuring, between what government delivers and what the public considers a basic level of competence, is widening from both ends at once. The system is straining to clear the old bar at the same moment the bar is rising. In a stable environment, H2- work that buys time for a failing system might be much-needed, and might be a missed opportunity for transformation. In an environment where disruptions of all kinds are accelerating, it becomes a compounding liability. Extending the lifespan of a brittle system just means the system eventually fails more spectacularly. More people get hurt. More people look for alternatives to democracy.

Sketching H2+

So what does H2+ look like? There certainly won’t be one definition everyone agrees on, but there are a few principles worth throwing out.

Move upstream. The most important marker of H2+ work is that it targets the conditions that produce the problem rather than the symptoms they generate. Getting a specific agency better digital services is H2- if it leaves untouched how digital projects are staffed, funded, and overseen across government. The upstream target is an operating model in which funding mechanisms, team structure, decision-making authority, roadmapping practices, and success metrics are all aligned around outcomes rather than outputs. A fellowship that places talented technologists in one agency and a civil service reform effort that changes how every agency can hire are both valuable. But they are not equivalent bets on the future of government capacity, and a portfolio that’s weighted heavily toward the former at the expense of the latter reflects a preference for the legible over the structural. H2+ philanthropy means being willing to fund work whose outcomes are diffuse, slow, and hard to attribute — because that’s often what structural change looks like from the outside.

Connect legislative and executive branch reform. The people working on congressional modernization and those working on executive branch reform have largely been operating in parallel, each with a partial picture. Legislative modernizers understand how the structure of congressional incentives shapes what agencies are asked to do and how they are overseen; executive branch reformers understand what those asks produce when they land in agencies. Together, that knowledge is considerably more powerful than either is alone. California’s Outcomes Review experiment — in which legislators are building structured feedback loops between the laws they pass and what those laws actually produce — is an early model of what it looks like when the legislative branch takes seriously its role not just in passing law but in learning whether law works.

Advocate, influence, build power. The reform field has been busy helping agencies navigate broken systems, building capacity to survive dysfunction, demonstrating that things could work better. That work was and is enormously valuable. But those needs will continue to grow as long as we put off structural reform. The people busy plugging holes in the system are exactly the people who most need to be in the reform conversations — they get what really needs to happen. It’s not about sidelining them. It’s about putting the pieces of the puzzle together and building a field that advocates effectively, and advocates for the right things. That means a diverse coalition whose messages can harmonize, including hard-hitting campaigners willing to play the game, not just technocratic analytical types.

Use federalism as a flywheel. States are laboratories of democracy5, but H2+ work doesn’t just wait for state and local governments to experiment and hope the results diffuse upward. It actively designs for the spread of proof points, using early-adopter states to generate evidence, building the connective tissue that carries lessons from those states to others and to the federal level and back down again. Progress up and down the federalism stack reinforces itself, if the field is organized to make that happen rather than leaving it to chance.

Subscribe now

The window

None of this means rescue work should stop, or that demonstrations are worthless, or that capacity substitution isn’t helpful and needed. Some H2- work, done deliberately and named honestly, is best understood as experimentation: we’re running it inside the failing system precisely because that’s where we’ll learn what a new operating model has to do. Both can be valuable.

But the field needs a shared frame clear-eyed enough to ask, with each investment: does this move the system toward H3, or does it prolong H1? That question should be driving how resources, talent, and attention get allocated now — not because the prior work was mistaken but because the moment is different and the cost of extending the status quo is too high.

Some things haven’t changed. The community is still full of good, smart people with enormous insight into a very difficult problem. We’ve just run out of time to do it the way we’ve been doing it. Every H2- intervention that returns the system to “good enough” is now a bet that good enough will hold. It’s a bet I no longer think we can afford to make.

The window for H2+ work has not been open like this before. It will not stay open indefinitely.

Share

I’ve disabled comments on this shorter version of the post. If you want to comment, go to the longer version and comment there.

But the government we have — the operating model it runs on, the rules and structures and assumptions that shape how it hires, procures, and delivers — was built for a world that no longer exists, and the distance between that world and this one is growing. We are approaching the kind of moment when that gap stops being a management problem and becomes a true legitimacy crisis. (Many will say that moment has already come.) It’s time to start asking whether the theory of change most of us have been operating under — incremental improvements off a pretty poor baseline — was ever going to get us to a government capable of meeting fast-changing needs. It hasn’t yet, and if we don’t do something differently, it won’t.

Kelly Born at the Packard Foundation recently shared with me a framework called the Three Horizons, originally developed by Anthony Hodgson and adapted widely in systems-change work. In it, Horizon 1 is the currently dominant system.1 It’s functional enough to persist but failing in critical ways, especially for people with less power. Horizon 3 is the future system you’re working toward, already visible in patches of practice that embody different values and different ways of working, but far from the norm.2 Horizon 2 is the turbulent middle where change agents work.

Subscribe now

But the key insight is that not all Horizon 2 work is the same. Some H2 innovations genuinely create the conditions for the new system to emerge. Call those transforming H2, or H2+. Others, however inadvertently, extend the lifespan of the failing system by relieving the pressure that might otherwise force structural change.3 Call those sustaining H2, or H2-. Both feel like reform, but they have very different long-term implications.

H2- work is attractive because it usually produces real value in the short run. H2+ work can take a long time to pay off, and the path is rarely clear. In a stable environment, you can get away with a lot of H2-. In an environment where the underlying system has become truly untenable, the difference between the two starts to matter a great deal. I think that’s where we are now.

A lot of what the government reform field has built over the past fifteen years, including a lot of my own work, is probably H2-. Not because the people doing it were wrong, and not because the work didn’t matter, but because the structure of the work, no matter how well executed, had a systematic tendency to solve the immediate problem in a way that made the structural problem easier to tolerate. If that sounds like an indictment, the first place I’d point the finger is at myself. And I don’t regret any of it. Nor am I saying that there’s not a valuable place for H2- work. There is. I’d just like us to see it for what it has been so we can be more deliberate about what we do next.

There are a few patterns that play out over and over in the category of H2-, the work that sustains the status quo.

The pressure valve

The dynamic of unintentionally propping up the status quo is easiest to see in what you might call vertical interventions. An agency lacks the talent to do important work, so philanthropy pays for some detailees, often through an Intergovernmental Personnel Agreement. (I was an IPA during my year in the White House.) An agency can’t hire the people it needs under standard civil service rules, so Congress grants a particular team a special hiring authority, as it did for the CHIPS implementation team. The procurement rules will take too long, so an agency gets an Other Transaction Authority exemption. The number of carve-outs Congress has granted to the Paperwork Reduction Act in recent years is its own story.4 The Abundance world is drowning in examples of exemptions to laws like the National Environmental Policy Act and the California Environmental Quality that have been necessary to get anything built, but have then left core barriers to the larger agenda intact. The answer to “the system doesn’t work” has been to hack it in the vertical domains that get attention, not fix it horizontally.5

Share

Each of these carve-outs delivers genuine value. Most of them are the right call in the moment. But each one also functions as a pressure valve on a system that, absent the relief, might generate the political and institutional pressure needed to fix the underlying problem. The legislative carveout that makes Paperwork Reduction Act compliance unnecessary for a particular program means the people and organizations with the most standing to demand PRA reform never do. They just move on to the next challenge. The rescue team that fixes a broken system well enough to make it functional returns it to “good enough,” which is exactly the condition in which structural change is hardest to achieve.

The problem isn’t the mechanisms themselves. IPAs and OTAs exist for good reasons and serve legitimate purposes. The problem is when they get deployed as substitutes for fixing the underlying dysfunction rather than as complements to work the system was never designed to do. There’s a difference between an IPA who brings in genuinely rare expertise that even a fit-for-purpose civil service would struggle to hire, and five IPAs filling roles the agency should be able to hire for itself if the system worked. Even the same tools can sometimes have opposite relationship to the structural problem.

The same dynamic shows up in the capacity conversation. When an agency lacks the staff to administer complex grants, the instinct (philanthropy’s instinct, as much as anyone’s) is to supplement that capacity with more staff, more technical assistance, more handholding through the process. But that instinct accepts the complexity as fixed. Louisiana’s Department of Education has been showing a different lesson in their education reform work: a cleaner, simpler grant framework needs fewer people to administer, produces better outcomes, and distributes resources more fairly because it’s actually navigable by grantees who don’t have grant compliance staff. The H2+ move is not to staff up to survive it, but to simplify the system,, and to strengthen the capability of that system.6 The instinct to reach for “government needs more resources” is understandable. It is also, at this point, a way of not asking the harder question.

The cost of the workaround or the staffing plus up isn’t just lost pressure. These short-term fixes have become enormously costly in their own right. Every special authority, every exemption, every vertical fix makes the overall system more complex, more fragmented, and more navigable by sophisticated actors — large contractors, well-resourced agencies, organizations with the staff to learn which door to knock on. (The way I put this in my book was “paperwork favors the powerful.”) The organizations least able to navigate that complexity are typically the ones serving the populations with the worst outcomes. Vertical interventions don’t just leave the underlying dysfunction in place. They tend to entrench it.

The demonstration that doesn’t scale

A close cousin is the successful pilot, a project that shows that SNAP enrollment can be simplified, agencies can build digital services in-house, or test-and-learn mechanisms can improve program outcomes, for example. The pilot works. But because the structural conditions for scaling don’t exist, it stays a demonstration, or it gets scaled in a form that loses the properties that made it work, because the procurement rules, the civil service rules, or the oversight structures were designed for a different kind of work. I suspect most readers who’ve worked in this space have a few of their own examples ready at hand, and I’d love to hear about those in the comments.

By scaling here, I don’t mean that a pilot program grows. Government operates at a very large scale, so a pilot program growing from, say, ten uses of a new hiring pathway to hundreds or even thousands is just a bigger workaround. The point isn’t that there’s now an alternative that some people can use with great effort. If the current system is broken, the point of the demonstration is to learn what could work to replace it, and then do just that. Scale means proof points turn into system reform.

In the Three Horizons frame, the existence of these pockets of better practice is real Horizon 3 evidence. They are signs of the future system we are looking for and trying to grow. It becomes H2- when the political system treats those pockets as proof that the problem is solved rather than as proof that the problem is structural, even as the people doing the work know how far from solved it is.7 Demonstrations that show something is possible but don’t change the conditions that make it the exception (especially in the four competencies I’ve talked about before) can end up providing the illusion of progress.

The philanthropic funding model tends to reinforce this pattern. Demonstrations are fundable because they produce legible, attributable, near-term outcomes — a number of people helped, a process improved, a tool deployed. Horizontal structural reform — changing the civil service rules, repealing the PRA, restructuring how digital projects get budgeted — produces outcomes that are diffuse, slow, and hard to attribute to any single funder or grantee. Philanthropy systematically underinvests in it because the incentive structures of grantmaking reward the visible and near-term. The result is a field well-resourced to demonstrate what’s possible and under resourced to change the conditions that keep the possible from becoming normal.

Share

The co-optation trap

The third pattern may be the most pernicious, and it’s the one I find personally most frustrating. The field promotes concepts that carry genuine Horizon 3 values. (Think user-centered design, evidence-based policy, feedback loops, etc.) And then the H1 system absorbs the vocabulary, where it does the work of legitimizing the status quo rather than challenging it.

“Agile” is now a procurement category that large contractors have learned to perform while delivering waterfall outcomes on agile timelines. “Human-centered design” is a workshop format and a contract deliverable. “Evidence-based” has become a requirement that agencies must demonstrate compliance with, which changes the relationship between agencies and evidence in ways that matter. When evidence production is mandated as a compliance activity, for example, it can crowd out evidence as a genuine decision-making tool. An agency that goes looking for the best possible data because it wants to know something is in a very different relationship with that data than an agency that produces a required report because OMB or Congress needed the box checked. From the outside the two can look almost identical. They produce very different organizational behavior.

The deepest version of co-optation is when the system doesn’t just absorb the vocabulary of reform, it learns to perform it. An agency can run user research, adopt agile ceremonies, and publish evidence reports while leaving the governance structures, vendor relationships, and incentive systems that produce bad outcomes entirely intact. The performance is often sincere. The people doing it believe in what the words mean. But copying the visible practices of a different way of working, without changing who has power, how decisions get made, and what anyone is accountable for, produces a somewhat nicer version of the same thing rather than a different thing. Reform becomes a ritual that signals values but doesn’t deliver on them. The system learns to speak the language, but it doesn’t actually change.

A different moment

The H2- work I’m describing has been done in good faith by people. I am one of those people. Code for America, which I founded and where I spent more than a decade, is in important respects capacity substitution. USDR, which I also helped start, is as well. The healthcare.gov rescue (which I didn’t actually work on but tried to provide moral support for) was the rescue-and-rebuild cycle. For much of the past fifteen years, the H2- path was arguably the right call. When there was no political space for structural change, demonstrations were a good way to build the evidence base and develop the field.

I think we are in a different moment now. This moment is defined by disruption. I count three kinds.

Contingent disruption — pandemics, climate events, geopolitical shocks, financial crises — is unpredictable in its specifics but very predictable in its category: large, fast-moving, high-stakes demands that fall disproportionately on government. COVID was not an anomaly. The next version won’t look the same.

The most recent disruption to federal government, however, was political. Whatever the cost of its methods, DOGE made the brittleness of the current operating model impossible to ignore and created political openings for structural arguments that previously had no traction. The reform field did not create this moment. But it can shape what comes out of it.

AI brings structural disruption. This is a transformation already underway in the material conditions of work, economy, and administration. AI creates dramatic change in both the needs and conditions government must respond to and the ways in which it can respond at the same time. Yes, I certainly mean a social safety net not nearly fit to handle the levels of unemployment that are likely coming our way, and yes, I mean possible upsets in the balance of power between agencies and the vendors they rely on, but that’s barely scratching the surface.

AI is not only an exogenous shock that government will have to absorb. It is also moving the bar on what counts as acceptable service in the first place. People are already using AI to understand their medical bills, navigate insurance denials, and draft appeals for benefits they were wrongly denied. Soon they will expect to apply for SNAP or file their taxes by uploading a paystub and answering a few plain-language questions, not by filling out even the best-designed web form. The forty-page PDF used to feel intolerable. The well-designed web form will start to feel that way too, and faster than the last transition did.8

And service delivery is only the most visible piece. The same expectation shift is going to hit regulation, permitting, enforcement, how quickly an agency can respond to a new problem, how a legislature decides whether a law is working. If a small team with the right tools can map a regulatory regime in a week, the timelines we have now, in which rulemaking takes several years--or even multiple presidential terms--become indefensible. If an advocate can stress-test a policy against thousands of edge cases before it gets enacted, the standard for what counts as due diligence in lawmaking starts to move. The bar is rising on the whole surface of what government does, not just on the forms people fill out.

Subscribe now

Not everyone wants this shift to happen. Public sector unions have secured laws in several states forbidding the use of AI in service delivery, won contracts requiring union consent before autonomous vehicles can operate, and pushed legislation mandating staffing levels that the work no longer requires — as my colleagues Robert Gordon and Nick Bagley have documented. The concern for workers caught in this transition is legitimate. But blocking government’s transformation while the world around it moves on is not a strategy for protecting those workers. It exacerbates public frustration with government, weakens the case for investing in it9, and leaves the people who most depend on public services with a system increasingly unfit to serve them.

So the gap we have been measuring, between what government delivers and what the public considers a basic level of competence, is widening from both ends at once. The system is straining to clear the old bar at the same moment the bar is rising.

In this environment, the benefits systems that struggled to scale during COVID will be asked to scale again. The regulatory processes that can’t move quickly will be asked to respond to developments they weren’t designed to anticipate. The civil service system that can’t attract the people it needs now will need to attract people with skills that didn’t exist a decade ago.

If I had to pick, it’s AI that drives this disruptive moment. But I don’t have to pick. You could just as easily imagine climate shocks, or the next pandemic, or an escalation of the current war. Truly, some combination of all the above is not that unlikely. Reasonable people may disagree about the size and shape of the disruption AI will bring, but betting against disruption generally seems deeply unwise at the moment.10

If you buy that argument, then we must acknowledge that a reform field largely dedicated to H2- work is not what the moment calls for. In a stable environment, H2- work that buys time for a failing system might be much-needed, and might be a missed opportunity for transformation. In an environment where disruptions of all kinds are accelerating, it becomes a compounding liability. Extending the lifespan of a brittle system just means the system eventually fails more spectacularly. More people get hurt. More people look for alternatives to democracy.

That doesn’t mean we need to throw everything out and start over. For the reform ecosystem, it means existing actors need incentives to align their work toward structural transformation, new actors with adjacent expertise need to be welcomed into the fold (especially advocates and lobbyists, given how little influence muscle the field has today), and connections need to be made both upstream and downstream of where we’ve been focused. It means articulating competing H3 visions from a wide range of ideological and practical perspectives and debating them among, including the project that sparked this line of thinking, which Kelly funded and FAI and New America are currently working on. It means designing funding and partnership structures that reward structural ambition while staying grounded in meaningful near-term progress. Funders and grantees share responsibility for creating the conditions under which a diverse set of actors can aim higher by working together, and connecting the dots upstream.

For this to work, it can’t be a zero sum game. Government capacity is wildly neglected in philanthropy despite its high leverage. (Good luck naming an issue philanthropists care about that doesn’t benefit from increased government capacity.) Could the field stop doing some H2- work? Sure. That would free up some existing resources for more H2+ work, which has been too little of the field’s mindshare and resources to date. But that is not the path forward — it wouldn’t get us where we need to be. We need more resources, full stop. We need to make the case to philanthropy for greater investment in the entire field (that’s part of what Recoding America Fund is trying to do) and make the case to government leaders, including electeds, to invest in better plumbing, so that the investment in H2+ work isn’t coming at the expense of the essential life support.

Sketching H2+

So what does H2+ look like? There certainly won’t be one definition everyone agrees on, but there are a few principles I’ll throw out in the hopes of sparking much more discussion.

Move upstream. The most important marker of H2+ work is that it targets the conditions that produce the problem rather than the symptoms they generate. Getting a specific agency better digital services, for example, is H2- if it leaves untouched how digital projects are staffed, funded, and overseen across government. The upstream target is an operating model in which funding mechanisms, team structure, decision-making authority, roadmapping practices, and success metrics are all aligned around outcomes rather than outputs — what Ann Lewis and I have called the product operating model. This requires changes in both the executive and legislative branches. An agency can do everything right internally and still fail if its appropriations are structured around meaningless requirements rather than continuous improvement, or if oversight bodies evaluate it on spending compliance rather than user outcomes. Getting digital right, at scale, means changing the legislative and administrative structures that govern what that’s even allowed to look like. (For a bright spot, look to the FAFSA team, which has successfully parlayed a rescue effort into meaningful structural change, including shaping its own oversight environment. More on that here and here.)

Funders need to ask not just whether an investment does good but whether it changes the conditions under which good can be done at scale. A fellowship that places talented technologists in one agency and a civil service reform effort that changes how every agency can hire are both valuable. But they are not equivalent bets on the future of government capacity, and a portfolio that’s weighted heavily toward the former at the expense of the latter reflects a preference for the legible over the structural. H2+ philanthropy means being willing to fund work whose outcomes are diffuse, slow, and hard to attribute — because that’s often what structural change looks like from the outside.

Connect legislative and executive branch reform. The people working on congressional modernization and those working on executive branch reform have largely been operating in parallel, each with a partial picture. Legislative modernizers understand how the structure of congressional incentives shapes what agencies are asked to do and how they are overseen; executive branch reformers understand what those asks produce when they land in agencies. Together, that knowledge is considerably more powerful than either is alone. California’s Outcomes Review experiment — in which legislators are building structured feedback loops between the laws they pass and what those laws actually produce (and getting rewarded for it with good press, the currency of elected officials!) — is an early model of what it looks like when the legislative branch takes seriously its role not just in passing law but in learning whether law works. The H2+ version of this is a field that actively bridges these two worlds, helping each understand what the other sees, and building reform strategies that are legible and actionable to both.

Subscribe now

Advocate, influence, build power. I remember when a hill staffer called me years ago for advice on an issue. I gave her my take, and suggested a few others she could also talk to, and as we were wrapping up she said “You know, it’s a lot of work for me to tap the civic tech folks. The industry groups…they’re organized. They have a point of view. They’re here on the hill talking to us all the time. You folks….I have to figure out the ten people to call and track them down. And then figure out what you’re saying and how to translate it. You’re ceding the conversation to vested interests.”

The reason we weren’t educating lawmakers and policymakers then is that we were busy -– busy helping agencies navigate broken systems, building capacity to survive dysfunction, demonstrating that things could work better. That work was and is enormously valuable. And it remains underfunded. There could be far more of it and needs would remain unmet. But those needs will continue to grow (especially as AI widens the gap between the public and private sectors) as long as we put off structural reform. Ironically it’s people busy plugging holes in the system who most need to be in the reform conversations. They get what really needs to happen. It’s not about sidelining them. It’s about putting the pieces of the puzzle together and building a field that advocates effectively, and advocates for the right things.

Above I’m talking about my past work in civic tech. The good government ecosystem has been more advocacy-oriented, for sure. Folks like Jenny Mattingly at the Partnership for Public Service have been a consistent voice on the hill for the public interest for years now. But who is the Jenny Mattingley of any given state? And even in Washington, Jenny and the few others like her are fighting mightily but outgunned by more organized interests with deeper pockets. There’s a real opportunity to define an agenda and resource a diverse coalition of groups to whose messages can harmonize. That diversity needs to include hard-hitting campaigners willing to play the game, not technocratic analytical types. That’s how we really win.

Ground the work in the actual politics. Building off my previous point, structural reform happens when there are political actors, in governors’ offices and agency leadership, but equally in legislative chambers and committee assignments, who have the standing, the incentive, and the capacity to move it. H2+ work is attentive to where those actors are and what they need. In the current federal environment, that means understanding which officials have genuine reform mandates and what constraints they’re operating under, which committee chairs have oversight interests that align with structural change, and where executive action is available and durable versus where statutory change is actually required. In states, it means working with the specific dynamics of a given governor’s office or legislative leadership. The field has too often treated politics as noise to be ignored rather than a medium to be worked. H2+ work takes the politics seriously, including the politics of the legislature, which shapes not just what agencies are authorized to do but what they are incentivized to attempt.

Build coalitions that make previously niche issues salient. For most of the past two decades civil service reform was something occasionally gestured at by good government groups and researched by a small number of policy scholars. Today, you can at least declare it a priority without getting laughed out of the room. What’s changing is not the substance of the analysis but the coalition around it. There is finally a convergence of people across the ideological spectrum who recognize, for different reasons, that the public sector workforce system is producing outcomes no one wants. That convergence is making civil service reform a live political issue. Same with Paperwork Reduction Act reform. And the same dynamics can work for other issues. H2+ investment identifies the structural reforms that are ripe for similar coalition-building and builds the political infrastructure to make them move.

Use federalism as a flywheel. States are laboratories of democracy,11 but H2+ work doesn’t just wait for state and local governments to experiment and hope the results diffuse upward. It actively designs for the spread of proof points, using early-adopter states to generate evidence, building the connective tissue that carries lessons from those states to others and to the federal level and back down again. Progress up and down the federalism stack reinforces itself, if the field is organized to make that happen rather than leaving it to chance.

Doing different things differently

I realized early last year that while I’d spent the bulk of my career trying to drag government into the Internet Era, that work has to change now. We are entering a new era, and if those of us who fought the last fight don’t adapt to the conditions and expectations of this one, we’ll make exactly the mistake the people who resisted internet-era ways of working made. We’ll become the blockers — the ones holding on to old ways of working because that is what we are used to and that is what we are good at.

None of which means rescue work should stop, or that demonstrations are worthless, or that capacity substitution isn’t helpful and needed. Some H2- work, done deliberately and named honestly, is best understood as experimentation: we’re running it inside the failing system precisely because that’s where we’ll learn what a new operating model has to do. That’s a different kind of work from rescue that produces learning incidentally, but both can be valuable.

But the field needs a shared frame clear-eyed enough to ask, with each investment: does this move the system toward H3, or does it prolong H1? That question should be driving how resources, talent, and attention get allocated now, not because the prior work was mistaken but because the moment is different and the cost of extending the status quo is too high. There will have to be work that sustains the status quo, but what tradeoffs are we willing to make?

But insisting we ask the question does not mean that answering it is easy: there is no objective set of criteria that distinguishes one from the other. What may look like H2+ to some may seem like H2- to others, and part of that depends on your particular vision of that third horizon (more on that in the coming weeks.) Some may see work as contributing to a transformation, and therefore H2+, but towards an undesired H3 state. Grappling with how to answer this question is work we all need to be doing. (You’ll hear more in the months to come about how the Recoding America Fund, the organization I co-founded and chair, is going to be using this framework to guide its work.)

If this framework is sparking thoughts for you, and you want to contribute, a good place to start might be thinking and writing about your third horizon: What is a desired future state for American government? How would it need to work to meet the needs of the public in the world we are hurtling into? The point in the future you choose is up to you, and somewhat arbitrary. A third horizon will always be a moving target, in that the ability to continually adapt to changing circumstances without causing political collapse is part of the change we seek. But tell a story about an operating model that’s fit for purpose, and let’s start comparing notes.

Some things haven’t changed. The community is still full of good, smart people with enormous insight into a very difficult problem. We’ve just run out of time to do it the way we’ve been doing it. A brittle system that gets propped up through manageable shocks will eventually meet a shock it can’t survive, and we are moving into a period where the shocks are neither manageable nor hypothetical. Every H2- intervention that returns the system to “good enough” is now a bet that good enough will hold. It’s a bet I no longer think we can afford to make.

The window for H2+ work has not been open like this before. It will not stay open indefinitely.

Subscribe now

Subscribe now

Here’s what it said:

Goodhart’s Law is one of the deepest traps in organizational life, and the leaders who handle it best tend to share a few recognizable instincts.

They treat metrics as evidence, not truth. The most effective leaders use numbers to ask questions rather than answer them. A good metric tells you something is worth investigating; it doesn’t tell you what’s actually happening. They stay relentlessly curious about what the number is and isn’t capturing, and they push their teams to bring them the messy reality behind the data, not just the data itself.

They rotate and retire measures. Smart leaders know that any metric will eventually be gamed — not always maliciously, but inevitably, as people optimize for what’s measured. So they treat metrics as temporary tools rather than permanent fixtures. They’re willing to abandon a measure once it has served its diagnostic purpose or once it starts distorting behavior.

They invest heavily in qualitative feedback loops. The antidote to a corrupted metric is usually direct contact with reality — customer interviews, frontline observations, case reviews, unscripted conversations. Leaders who do this well build it into their routines structurally, not just as a crisis response. They go to where the work is actually happening.

They make it safe to report bad news and gaming. One of the underappreciated drivers of Goodhart’s Law is that people often know a metric is being gamed but won’t say so because the organization has implicitly rewarded hitting the number. The best leaders create explicit permission — even reward — for surfacing the gap between what’s measured and what’s real. This requires psychological safety, but it also requires leaders who visibly don’t shoot the messenger.

They hold multiple measures in tension. Single metrics are the most vulnerable to Goodhart’s Law. Sophisticated leaders use portfolios of measures that push against each other — speed and quality, output and outcomes, short-term and long-term — so that gaming one comes at the expense of another. The goal is to make it harder to optimize for the measure rather than the underlying thing.

They articulate purpose clearly enough that the metric becomes less important. This is perhaps the deepest solution: when people genuinely understand what they’re trying to accomplish and why it matters, they’re more likely to flag when the metric is pointing in the wrong direction. Mission clarity doesn’t eliminate Goodhart’s, but it changes the culture in which it operates.

The leaders who fail at this are usually the ones who respond to metric failure by adding more metrics or tightening accountability around the existing ones. The leaders who succeed understand that the problem is structural — that measurement and reality inevitably diverge — and they build organizations that can navigate that gap rather than pretend it doesn’t exist.

Share

I shared this with Dave, and he thought it was pretty good too. The only problem now is that we need to find something else to fight about. I’m not terribly worried about that, though.

I hope you find this useful. Please share what you’ve seen work in the comments.

The problem is that no single agency’s budget can fix this. These shared systems — digital identity, secure data exchange, real-time payments infrastructure — aren’t anyone’s capability. They’re foundational inputs to every agency’s capabilities. In a new article in Finance & Development Magazine, David, Diane Coyle, and Beatriz Vasconcellos argue that they should be thought of, funded, and governed the way we think about roads and power grids: as infrastructure. They should be long-lived, cross-cutting, publicly governed, and designed for reuse across sectors and agencies and even across time.

Subscribe now

It’s telling that this article is aimed at finance ministries around the world, not digital leaders. Digital ministries typically know what shared infrastructure is needed, but they lack the budget authority and cross-governmental leverage to actually build and sustain it. (To be fair, they are not consistently fans of DPI. If they’ve spent their careers trapped in the dysfunctions of legacy government IT, they may not trust that a central office can deliver something that will work for their particular needs, and proving that it can can be a bit of a chicken and egg problem.) But finance ministries have that authority. What they’ve mostly lacked is the framework — and the willingness — to use it. A finance official they interviewed put the structural problem exactly right: “There’s no category in the budget for systems that benefit everyone but belong to no one.”

Of course, that’s why here in the US we have GSA, and particularly the Technology Transformation Service, home of still nascent (after many years) cross-government platforms like login.gov. It’s been a long, hard road for login. I won’t go into its epic history here, but the news that TTS will now report into federal CIO Greg Barbaccia (or rather, he’ll now be dual-hatted as acting TTS director) gives me hope. Having TTS separate from OFCIO has always made a coherent strategy for infrastructure challenging. I think Barbaccia is serious about leapfrogging digital into the current era, and the fact that he’s committed to investing in login.gov is a sign he’s taking a practical, thoughtful approach. May Congress trust his leadership and budget appropriately for our “systems that benefit everyone but belong to no one.

But Congress also needs to learn this same lesson at the agency level. David’s article reminded me of the great work by Solitaire Carroll, who pioneered something called capability-based budgeting during her time at the Department of Veterans Affairs, and has written about this for the Niskanen Center. The idea sounds technical, and the implementation absolutely is, but the underlying diagnosis is simple: the federal government funds technology by tying dollars to systems, programs, and organizational boxes on a chart. If Congress appropriates money for a claims processing system, that’s what the money is for: that specific system. Not “the VA’s ability to process claims,” which might involve a dozen systems, some legacy, some modern, some in between, plus the people who maintain them and the teams who improve them. Just this one bucket, with this one label, which a given oversight committee may have decided should exist for its own reasons.

The result is that the actual capability — the thing veterans need — can’t be funded or managed as a coherent whole. You can’t ask “how much does it cost us to process a claim?” in any meaningful way. You can’t make sensible tradeoffs between investing in a legacy system and replacing it, because the dollars aren’t organized to allow that question. The money is locked into boxes, and the boxes are organized around org charts and appropriations history rather than around what government is actually supposed to do.

Solitaire’s solution — tested at VA with hard-won real results — is to reorganize IT investment around enduring capabilities: eligibility and enrollment, claims processing, benefits delivery, and so on. You identify what the agency is responsible for delivering, group all the systems that contribute to that delivery, and start asking “how much for claims processing?” instead of “how much for System X?” Capability-based budgeting is product funding operationalized. The two concepts say the same thing: we need to fund ongoing capabilities, not static things. What Solitaire has done is figure out how to actually do that inside a federal budget process that’s not designed for it.

If it sounds straightforward. It is not. It requires convincing appropriators, OMB, and internal stakeholders to think differently about categories of spending they’ve been managing the same way for decades. Solitaire’s work at the VA involved not only lots of education of a wide range of stakeholders, but also astonishing feats of reprogramming — complicated, detailed work to get money appropriated in one mindset to function in another, and serve the purpose it should. This is thousands of hours of work that should not be necessary. But at least this reprogramming is possible, and the VA case study shows it.

We’re going to have a lot of work getting appropriators, OMB, and internal stakeholders to think differently in the coming years. Carroll’s reform can in principle be driven by a single motivated CIO working within existing budget structures. It could spread from agency to agency as a best practice. But I think we’ve run out of time for the kind of retail approach, given the range and scale of disruption government must now contend with: workforce changes, budget constraints, and, of course, AI —not just the mandate to adopt it for public sector use, but the way it will shape the needs government must meet. (We might also mention crises like the war in Iran, which will dramatically increase the national security needs the DoW must meet.) What the Coyle/Eaves paper calls for isn’t so much a management reform, but a political and institutional one. In an international context, they are calling for finance ministries to proactively coordinate across independent agencies — to assert a strategic, cross-governmental role in digital investment that most of them have never played. Here in the US, we have the seeds of that, but there is enormous work to change the conditions under which platforms like login operate. That will look like specific reforms of the kind Solitaire and David call for, but underpinning those reforms is the challenge of getting the people who write the checks and hand out the grades to think differently about their roles and adopt a new vision of what good looks like.

What changes about their roles? For one, how they evaluate investments. The value of shared digital infrastructure comes precisely from the fact that it enables things beyond any single agency’s mission — including things in the private sector, things we can’t fully predict in advance, things that don’t show up in any cost-benefit analysis conducted at the point of investment. Malawi improved its credit markets by introducing a biometric identity system because lenders could suddenly verify borrowers more reliably. That outcome wasn’t in anyone’s project justification. Standard investment appraisal, Coyle and Eaves argue, is simply the wrong tool for evaluating these investments — built for physical infrastructure, it misses spillovers, long-term market effects, and distributional consequences almost by design.

But deeper than that, those who hold the purse strings must grapple with the misconception that governing is fundamentally about mandates and constraints. When I am asked for my advice on how to make an underperforming agency or team do better, I am usually being asked for mandates and constraints someone in a position of authority (exec or leg branch) can impose on an agency. It is the wrong question. These implementers are groaning under the weight of decades, even centuries, of mandates and constraints that have always been added, and never subtracted. We (the public, Congress, watchdogs) want agencies to do things, but have created an operating environment in which it is a miracle they can do anything at all. Adding more mandates and constraints only makes it worse.

Share

But there is a different way to approach the problem of agency effectiveness. Instead of adding, subtract. What are they doing today that they no longer need to do? What is hard for them to do that you could make easier by removing constraints? If you want an agency that’s responsive to the needs of the moment, you have to work in an enablement and capacity building framework. Investment in common infrastructure that will benefit them is another form of enablement and capacity building.

This idea gives lots of leaders the willies. It sounds like a loss of control, where control over the agency is the point. But they also know that the control they seek is an illusion today; they are frustrated that what they ask for and what they get are two very different things. Maybe the way to think about it is to return to the metaphor of a Rube Goldberg machine, which I’ve invoked frequently. Every time you add another mechanism, it can feel like added control, but if you pull back, you’ve created a system so complex, fragile, and bespoke that it’s a miracle when it works. Troubleshooting these systems becomes exhausting. If you want to put a nail in the wall, maybe just use a hammer directly.

Accordingly, if you want to verify who someone is online in a reliable, secure, low-friction way to enable almost everything government does digitally (benefits eligibility, healthcare access, licensing, tax collection, procurement) maybe just have a fit-for-purpose digital identity platform. If you want to the VA to be able process claims, maybe just give them money to build that capability. “Just” here doesn’t mean without reasonable controls and accountability, but it does mean getting back to basics and shedding some of the unnecessary cruft, complexity, and indirection. Requiring agencies to budget separately for building a system versus operating and maintaining (and not use money intended for one for the other) it is a great example of a control that entirely backfires and should be put out to pasture. There are many others.

We need new infrastructure if we want a government capable of meeting our country’s needs – needs that are accelerating and changing as the world around us experiences disruption of many kinds. That infrastructure will come in the form of common platforms, as Beatriz, David, and Diane suggest, and in the form of common sense capability-based budgeting and appropriations, as Solitaire has shown. But it will also come in the form of reformers working upstream of their previous efforts, changing the outcomes we get by changing the way we budget for and oversee investments. And it will come in the form of an evolving understanding of what governing needs to look like in the second quarter of the twenty-first century. The problems we face cannot be solved with the thinking that created them. Investing in that new thinking will be the most profound infrastructure investment we could make.

Share

Subscribe now

I’m interested in how that need is being met, and what we can learn from it. One place to look is Louisiana’s Department of Education, the only state education agency in the country to have achieved gains in both fourth-grade reading and mathematics on the National Assessment of Educational Progress between 2017 and 2024.1 This is the kind of meaningful outcome that is the true measure of state capacity. It’s not the number of staff or the size of the budget, but evidence of the thing people want out of government. In this case, it’s schools that do a better job of educating our kids. In fact, the Louisiana’s DoE story shows one awkward truth the DOGE-hopefuls instinctively knew: that sometimes less money and less staff actually do help create the conditions for better outcomes.

But it also shows how wildly off the mark the DOGE approach was. To paraphrase John Kamensky’s line about Reinventing Government’s shortcomings, you can’t eat dessert first. (He meant you should first cut the work by rightsizing procedure and streamlining priorities, then cut the workforce.) Better government is possible, but it’s hard work. Thoughtful frameworks, persistently and responsibly applied, help.

Share

I’m flattered that in Julia Kaufman and Kunjan Narechania’s telling of the Louisiana story, that framework is one Andrew Greenway and I described in our Niskanen paper “The How We Need Now.” I want to be crystal clear that all the work Julia, Kunjan, and many others did happened long before our paper was written. If there are parallels, you could almost say they invented the framework and we wrote it down, entirely unaware of each other. Kaufman and Narechania map LDoE’s four key actions directly onto the four pillars Andrew and I described: the right people focused on the right work, with purpose-fit systems (in Louisiana’s case, the modest but effective “Super App”) and test-and-learn frameworks (feedback loops and classroom observation). The parallels are close enough that reading their paper is a bit uncanny. But they’ve also extended the framework in at least one important way, and identified a gap in it that I want to address.

The road to incoherence is paved with good intentions

What Louisiana’s Department of Education accomplished is impressive. The state is tied for the highest child poverty rate in the nation, 39th in per-pupil spending. But when the Urban Institute adjusted the 2024 NAEP results for student demographics, Louisiana ranked second in the nation in both fourth and eighth grade reading.

To understand how they got there, let’s start with a few decades back.

State departments of education have been accreting complexity since 1965, when the Elementary and Secondary Education Act dramatically expanded federal funding for K-12 schools. More federal money meant more federal programs, and more programs meant more offices: a Title I office for schools serving low-income families, a Title II office for teacher quality and professional development, a Title III office for English language learners, an IDEA office for special education, etc. State agencies mirrored this structure, building parallel offices to manage each funding stream, each office with its own requirements, its own monitoring apparatus, its own theory of what districts should be doing. When No Child Left Behind arrived in 2001, it layered a performance management apparatus on top of the compliance infrastructure that already existed, rather than replacing it. By the time John White became Louisiana’s State Superintendent in 2011, the state had both systems running simultaneously, largely disconnected from each other.

What this meant in practice was that most Louisiana districts completed more than 50 separate grant applications to receive state funding, each submitted to a different office, each with slightly different goals. Literacy targets, for instance, might appear in applications for Title I, Title II, bilingual education, and special education funding, with each framed differently, and each monitored by a different team. As one state official put it, there were 600 people on staff and about 300 of them were emailing districts about something. This is incapacity masquerading as capacity. You have the people, but what they’re charged with doing is collectively incoherent.

As with most dysfunctional bureaucracies, the vast majority of the people were doing a great job at the jobs they were asked to do. Each office was executing its mandate. The Title I team was doing Title I. The literacy office was pursuing literacy goals. Each was doing so thoroughly and diligently. But they were doing the wrong jobs. The dysfunction wasn’t a product of incompetence or bad faith. It was the predictable output of a structure that had been built incrementally, program by program, the result of decades of well-intentioned federal legislation.

The implementation chain keeps it together

What LDOE did starting in 2012 included cutting the workforce roughly in half, from 600 to about 300. State officials later noted that the central office workforce reduction was “helpful to coherence because they created an environment where there just weren’t as many people out there saying what they wanted to say.” So: workforce reduction, positive effect on outcomes.

But that sentence gets the causation backwards if you read it in isolation. The cut didn’t produce the coherence. The coherence made clear which positions were producing noise rather than signal, and then the cut followed.

Before any of the restructuring, LDOE did something that sounds simple and proved to be extraordinarily difficult. They defined what they actually wanted to see happen in classrooms in ways that everyone could understand and contribute to. The goal wasn’t “raise test scores” but rather the specific student experiences that would have to be present before test scores could rise. In reading, it was students engaging with complex texts and writing about them. In math, it was students working through problems and justifying their thinking. These goals were observable, measurable, and specific.

(By the way, I see a lot of plans for improving customer interactions that would benefit from this kind of specificity. “World class customer experience” isn’t actionable. Describing concretely what your users should see, do and feel in their encounters with government helps teams know what to actually change.)

From that vision, they worked backwards through what Kaufman and Narechania call an “implementation chain”: what would teachers need to do? What would principals need to do to support teachers? What would districts need to do to support principals? What would the state need to do to support districts? They named every link and specified every that would be required.

Subscribe now

This is where Kaufman and Narechania make their most valuable contribution to the framework Andrew and I laid out. We emphasized the test-and-learn approach — iterate, measure, adjust — but we didn’t fully articulate what has to be true before you can test and learn productively. Their answer is the implementation chain: a concrete, shared map of what every actor in the delivery system needs to do differently, derived from a precise definition of what success looks like for the end user. Without that map, you’re not testing toward anything in particular. You’re just iterating in the dark.

Only then did the restructuring follow, and it followed directly from the chain. They didn’t reorganize around org-chart logic or budget lines. They reorganized around three functions that the implementation chain required: a planning team to build and communicate the vision, a funding team to consolidate all resources behind it, and an implementation team to go into schools and verify that the chain was actually working. The 50+ grant applications became one “Super App” tied to a single unified strategy.

The workforce reduction wasn’t the reform. It was the residue of the reform. Similarly, the Super App wasn’t the reform. It was a tool that enabled the reform.

The How needs a clear What

This case also surfaces something that I think has occasionally been misread about the capacity framework. The framework is explicitly about the How. The best policy is just words on a page unless there is a way to have it make real impact in the world. That’s the how. Andrew and I care about this distinction (and talked about it in our paper) because we think that government needs a strong implementation muscle even when the folks you didn’t vote for are in power. “State capacity for me, but not for thee” doesn’t work. It results in capacity for no one. No one can get anything done, government fails the public, and it’s one more nail in democracy’s coffin. So the how matters, independent of the what.

But what we didn’t quite say is that there’s one critical thing about the what. You have to define clear goals. Louisiana’s story shows that you can’t build an implementation chain if you don’t know what you’re trying to implement. You can’t reorganize around a mission if the mission is fuzzy. The test-and-learn approach is only as useful as the clarity of what you’re testing toward.

I’m not a sailor, but I do remember being taught about tacking as a kid. To sail into the wind, you tack back and forth across it, which looks indirect but gets you there faster than heading straight into the headwind. The How is the tacking. It’s a lot like the build-measure-learn cycle, because you’re constantly adjusting. But if you don’t fix clearly on the point on the shore where you want to land, all that back-and-forth just gets you lost. LDOE fixed their point on the shore — students reading complex texts, students working through math problems — before they tacked. That’s what made the tacking work. I regret that Andrew and I weren’t clearer about this in our paper, and I’m grateful to Kaufman and Narechania for making it so explicit.

Share

Make the right choice the easy choice

One more thing LDOE did that deserves attention: they understood that signaling quality isn’t enough if adopting quality is hard.

They replaced the old curriculum approval process with a transparent tiered review, and published it. They negotiated statewide contracts so districts could buy Tier I materials at reduced cost without going through their own procurement processes. They built professional development opportunities that were only useful if you were using the good materials. They eventually required low-performing districts to use Tier I materials as a condition of funding, but only after enough districts had voluntarily adopted them that there were local advocates for the change.

This is choice architecture, and it’s the opposite of mandate-by-fiat. It assumes that most people in the system want to do the right thing and need the friction removed, not a new rule added. It assumes that the implementer closest to the work — the teacher, the principal, the district curriculum director — is an asset to be supported, not a variable to be controlled.

It’s worth noting that Louisiana’s neighbors have also seen reading score improvements in recent years, and much of the attention has focused on a single change: the shift from whole language learning to phonics-based reading instruction. States that have moved toward evidence-based curriculum have generally seen results. Louisiana moved early and decisively on curriculum quality, and that’s part of the story here.

But only part. Louisiana didn’t just order curriculum change. They built a system that could actually get that teaching into classrooms consistently, at scale, across a state with significant poverty and limited resources. Other states now, including my former home state of California, are following suit. There, for example, Assembly Bill 1454, signed by Governor Newsom last October, is trying to turn California into a Louisiana. It requires the State Board of Education to produce an approved curriculum list of materials aligned with evidence-based practices, funds teacher professional development aligned to those materials, and updates administrator credentialing standards to include literacy instruction support. But it doesn’t address the internal structure of the California Department of Education — the siloed offices, the fragmented funding streams, the incoherent messages to districts. That work doesn’t need to be done by the legislature, but it should be a key priority of the new administration, and Californians should be looking for a leader who gets this when they go to the polls in November. If California or any other state wants to see real results, they can’t stop at the “signal quality” and “align professional development” pieces. They’ll have to reorganize, consolidate grant applications, and build the feedback loops of implementation teams in classrooms four days a week.

Same number, different math

It’s meaningful (to me, at least) that the LDOE cut its central office staff by half. I’ve talked about 50% workforce cuts before. A senior Air Force official once told me that the only way to force meaningful change at the Pentagon would be cuts of around half the budget (I didn’t disagree, but if that ever happens, I hope they talk to Louisiana). And before someone made it clear to him that this was wildly impractical, Vivek Ramaswamy proposed firing every federal worker whose SSN ended in an odd number – again, half. These stories look similar. But they are very different.

DOGE’s theory was that the incoherence was the product of too many people, and that removing people would remove the incoherence. LDOE saw the incoherence as the product of a structure that pointed too many people in different directions. Removing the structure and replacing it with something coherent made some of those people redundant. That’s the correct order of operations.

Subscribe now

But where did that incoherence come from in the first place? Baton Rouge didn’t come up with that tangle of programs and incentives and reporting and metrics, Washington did. What LDOE previously did to districts, the federal Department of Education has long done to states. The structure of federal education funding, with its separate offices and siloed requirements, is what state agencies mirrored when they built their own siloed structures. And the great irony is that one more place that’s seen its workforce cut by half is the federal Department of Education.

Sadly, these cuts aren’t likely to bring the kind of coherence that would help every state, the way LDOE’s helped every district. The Trump administration ate dessert first. They are parceling programs out to other agencies, some with little relevant expertise, as part of what the department itself calls its “final mission.” The Office of Elementary and Secondary Education, which oversees Title I and the Every Student Succeeds Act, has seen cuts across all of its teams. There is no implementation chain. There is no vision of what students should experience in classrooms. There is no Super App consolidating incoherent funding streams into a unified strategy. There is just the cut.

If any administration — current or future — wants to see Louisiana’s results replicated nationally, it should look honestly at its own role in creating the conditions that made coherence so hard to achieve in the first place. That means that while the cuts at the federal Department of Education were the wrong move, simply replacing all those people would also likely be the wrong move. You don’t recover from DOGE with more of what made so many people want them to succeed in the first place.

There is no magic wand. The work takes years and requires, as Julia and Kunjan explain, buy-in at every level, from classroom teachers to state legislators. It’s not that reformers are timid or captured by the institutions they’re trying to change, certainly not in this case. But you are asking everyone in a large system to shift their daily behavior toward a new shared goal, and that requires understanding the goal, trusting that the people asking you to change have thought it through, and having the support to make the change. None of that can be done from afar. And absent a crisis, it takes time. (Never count out crises to speed up alignment, though.)

The need that drove people toward DOGE hasn’t been met. But the answer isn’t simply more staff and more complexity either. Those can’t be our only choices: unmanageable bloat, or arbitrary cuts. Luckily, they’re not. I’m sure Louisiana schools are far from perfect, but if the rest of the country was seeing gains like these, we’d be in a very different place. Maybe if people pay attention to what it takes to achieve them, we will.

Share

Today we know what he was doing. The escape fire consumed the fuel around him, allowing the main fire to pass over him and a few of his colleagues. But in 1949, the families of the 13 other smokejumpers who died accused Wag of causing their deaths. To them, what he had done made no sense.

Subscribe now

I love that Marina Nitze, Matthew Weaver, and Mikey Dickerson chose this story as a framing device for their new book, Crisis Engineering: Time-Tested Tools for Turning Chaos Into Clarity, out today. Not just because it brought back the memory of a book that I once loved, but because Maclean’s obsessive investigation of what had happened back then (he wrote the book years after the incident) seemed to me almost as heroic as the bravery of the smokejumpers. And indeed, his insistence on making sense of what happened has probably saved lives. Escape fires are now formally recognized and taught as a last resort tactic when training new firefighters.

The Dodge escape fire wouldn’t seem to have much to do with Three Mile Island or healthcare.gov or the pandemic unemployment insurance backlogs, but the authors use it to make a point about how action and understanding interact in a crisis. One key is exactly what Maclean himself did so well: sensemaking. In a crisis like Mann Gulch, sensemaking disintegrates: a broken radio, wind so strong communication is impossible, fire whose behavior violates well-tested assumptions, and a team scattered. You don’t achieve sensemaking by staring at a map; you achieve it by acting and observing results. Wag Dodge didn’t understand fire behavior well enough to explain the escape fire in advance. But his actions created the understanding itself — retrospectively, as all real sensemaking is.

The book’s key claim is that crises are opportunities, and the authors leverage Daniel Kahneman’s Thinking, Fast and Slow to explain why crises are the only real windows for organizational change — and why everything else, the incentives, the logical arguments, the reorganizations, mostly doesn’t work. Most organizations, most of the time, run on autopilot. People habituate to their environment, rationalize away small surprises, and build stable stories about how things work. A crisis breaks this. When surprise accumulates faster than the brain’s “surprise-removing machinery” can rationalize it away, the whole apparatus jams, and organizations become, briefly, reprogrammable.

An institution resolves a crisis in one of three ways, according to the authors. It makes durable deliberate change, it dies, or, most commonly, it rationalizes the failure into an accepted new normal. “Most large organizations contain programs and departments that passively accept abject failure: infinitely long backlogs, hospitals that kill patients, devastating school closures that do little to affect a pandemic. These are fossils of past crises where the organization failed to adapt.”

Too many of our public institutions have failed to adapt, and the idea that they might be reprogrammable at all is a bit radical. We live in an era when too many people have given up on them, willing to burn them to the ground rather than renovate them. If crises represent the chance for true transformation, then we’d better get a lot better at using them for that. This is explicitly why Crisis Engineering exists, and it’s a detailed, practical book — the theory and framing devices are well used, but there’s a ton of pragmatic substance here you’ll be grateful for when the moment comes.

Share

I remember when I was working in the White House and frustrated by the slow pace of progress. My UK mentor Mike Bracken told me: Hold on, you just need a crisis. You Americans only ever change in crisis. Boom. About two months later, healthcare.gov had its inauspicious start. And he was right. Change followed. Not all the change we needed, but a start. Marina, Weaver, and Mikey are three of the people who drove that change. I got to work with them again the first summer of the pandemic on California’s unemployment insurance claims backlog. I’m not a crisis engineer, but their strategies and tactics have deeply influenced how I think about the work I do and how I think we’re going to get from the institutions we have today to the ones we need.

We may be living in an era when too many people have given up on institutions, but we are also likely entering an era of crisis, and even polycrisis. This makes for uncomfortable math, but also drives home the need for a new generation of crisis engineers.

When I first read about Mann Gulch, so many years ago, I remember being in awe of the ingenuity and courage it took to start Wag Dodge’s escape fire. Today I think a lot about that pattern: the controlled burns that reduce the risk of megafires, the little earthquakes that take the pressure off faults under great tension, the managed crises that, if we’re skilled enough to use them, keep our institutions from the kind of collapse that comes when nothing has been allowed to give for too long. Dodge didn’t burn things down. He burned a path through. We’re going to have to get good at that.

Subscribe now

Subscribe now

My answer was that people expect to be able to use the internet to interact with their government. For a growing number of people, the internet was how they accessed almost every other service in their lives. When dealing with government meant in-person visits during working hours, and confusing, sometimes insulting paper-based forms, it not only cost people time they didn’t have, it also caused them to miss out on benefits they needed, to get in legal tangles when they failed to pay a ticket or comply with a regulation, and to generally feel that government was the enemy. These dynamics were clearly eroding trust in public institutions.

Later, I encountered Tom Loosemore’s definition of digital: “Applying the culture, processes, business models and technologies of the internet era to respond to people’s raised expectations.” I recognized in that definition the instinct behind Code for America: design government services around people’s real needs as those needs evolve, not around the convenience and habits of institutions. This was the service delivery version of Marci Harris’s “pacing problem.” Science and technology move fast, but government moves slowly, and this is now the source of a deep and fundamental problem in our society.

I share this as background to my take on an active discussion of the role of AI in government, spurred most recently by Erie Meyer’s piece in FedScoop on Friday. I love, admire, and deeply respect Erie, and I think she raises some valid concerns about philanthropic funders specifying AI as a condition of funding for government-related work. We absolutely should worry about hype cycles, misplaced priorities, and technosolutionism. When the emerging technology of the day was the blockchain, I shared with many in civic tech a contempt for that “solution in search of a problem.” But with AI, I think Erie and many others in this community are misreading key aspects of the need, the opportunity, and the role of philanthropy.

Share

Raised expectations, again

People are already using AI to understand their medical bills, navigate insurance denials, draft appeals for benefits they were wrongly denied, and to parse lease agreements and court filings written in language no layperson was ever expected to understand. This tax season, Claude is scoring a lot of points by finding savings for people as they file their taxes. There’s been a lot of concern about how these tools might lead them astray, but folks like Dave Guarino are methodically testing how well they do at helping people navigate the complexity of programs like SNAP. Instead of warning that models could prompt you to claim a tax deduction that’s no longer legal, consumer advocates should be testing to what extent they actually do. The news from Dave on SNAP, for example, is that they’ve gotten quite good.

For years, we celebrated the transition from a forty-page PDF (or a barely disguised web version of the same) to an intuitive, well-designed web form that had been informed by proper user research—and rightly so. But expectations are changing again, even as there remain hundreds of thousands of forms (online and paper) that still need redesigning. Soon, people may expect to have a plain-language conversation with a system that asks questions in ways they understand and does the hard work for them. They’ll expect to apply for SNAP or file their taxes by uploading a paystub and letting the AI extract the data, calculate eligibility, and ask only the questions it can’t already answer and only the ones actually relevant to their situation. And they’ll expect it to ask those questions in ways that they understand. At some point, even a very well-designed web form will start to feel the way the forty-page PDF used to feel. And that shift may happen faster than the transition to internet-enabled services did.

Luukas Ilves and his co-authors make this point well in a paper from last year called the Agentic State. Luukas describes a future in which “public services flow directly around people’s lives.” Needs are anticipated and solutions assembled across agencies without the user ever having to understand which department does what. But the future is already here. It’s just, as William Gibson said, unevenly distributed. Ukraine has a national AI agent that lets you describe a need in plain language and the agent completes the workflow end-to-end, rather than routing you through forms. Imagine a US state doing this.

There’s a dark side of AI’s arrival that we have grapple with, too. It’s not only people seeking benefits or navigating permits who will use AI. AI-enabled fraud in unemployment insurance and other benefit programs has grown substantially in recent years, and the attack surface is expanding. Luukas is correct about the stakes. Governments that fail to develop their own agentic capabilities will find that bad actors “run circles around public administration, using agentic capabilities to achieve their goals.” The choice, as he frames it, isn’t between an AI-equipped government and a simpler, more human-scaled one. It’s between a government that develops its own capacity and one that gets outrun by the people exploiting its absence.

Subscribe now

And AI can equip government for far more than benefits enrollment and fighting fraud. It is uniquely suited to the underlying problem that plagues government services: crippling policy cruft. When New Jersey’s Labor Commissioner Robert Asaro-Angelo was called to testify about his state’s unemployment insurance backlog, he brought along several boxes labeled “7,119 pages of active UI regulations” and set them in full view of the legislators. The technological issues his agency faced were a symptom. The cause was the sheer volume of rules that computer systems were supposed to operationalize. Agency guidance layered on regulations layered on statutes, each adding conditions and exceptions and cross-references, until the result is functionally impossible for an ordinary person, or even an ordinary caseworker, to navigate.

The field of civic tech has for too long accepted this complexity as a fixed constraint, in part, of course, because of the unhelpful norms that tell tech people to “get back in their lane” if they venture into policy. An AI-forward civic tech field would see enormous opportunity to change that. Just as the volume of rules is reaching entirely unsustainable levels, the tools to untangle and thoughtfully simplify them have arrived. Stanford RegLab’s Statutory Research Assistant (STARA) and other commercial options like those from Vulcan Technologies can read an entire regulatory regime, identify redundant, conflicting, vestigial, and simply unhelpful provisions, and return a map of the tangle that no human team is likely to produce. These tools can help us weigh tradeoffs, so we move away from frameworks that are the sum total of every conceivable idea that sounded good in the abstract but in aggregate make policy unimplementable at scale. And mapping the problem is just the first step. The Agentic State paper points toward something more ambitious: encoding rules as executable logic rather than ambiguous prose, allowing policymakers (and advocates, one would hope) to simulate how a proposed regulation would actually perform before enacting it. This means stress-testing effects, surfacing unintended consequences, and identifying the edge cases that turn well-intentioned rules into precisely the kind of tangle Commissioner Asaro-Angelo hauled into that hearing room. This work still relies on human expertise and judgement, but this is incredibly hard work that AI tools, and norms like rules as code, could make far easier.

AI can’t do anything about the political will to actually unwind and simplify these policies, regulations, and laws, but we haven’t really tested our leaders on this front. The way to find out if our law- and policymakers are willing to do the hard work of tidying up these messes is to hand them options, and see if they’re willing to make much-needed change. The pull of the status quo will be strong, so we should be relentless in teeing up opportunities for leaders to truly lead by using AI to rightsize these statutory and regulatory frameworks. If we use AI merely to navigate them, it will serve to excuse and enable further bloat. When benefits systems fail during times of need, it’s not the fault of the caseworkers, and it’s not even really the fault of the COBOL code everyone loves to blame. It’s the 7,119 pages. AI gives us a new set of tools to actually address the problem, if we choose to use them.

(This is just one example of where I hope civic tech advocates move as our world evolves: upstream. We must change the conditions under which technology in government is built and bought, not just the technology itself. Policy complexity is one of those crippling conditions, but there are others.)

AI has also given us a new set of tools to improve access to government services, especially for non-English speakers. US Digital Response, for example, has had enormous success helping governments use AI for translation. In New Jersey, so many more claims came in from ESL applicants after the translation work that the Commissioner (same one as pictured above!) was reportedly upset, realizing how many people must not have been applying before. This wouldn’t have happened without AI, but it also would have been unlikely to happen, at least when it did, without philanthropy investing in USDR to help the state and many other governments do the same. (See my point about philanthropy as catalyst at the end.)

Share

The vendor argument runs both ways

Erie understandably worries that encouraging government to use AI tools sold by tech companies will deepen dependency and entrench the current dynamic. If “use AI” always means “buy an AI product from a large vendor and integrate it into your existing systems,” we could reproduce our current problems: more lock-in, more opacity, more contracts that agencies can’t escape. But that is a distinction her piece doesn’t quite make, and it’s an important one.

Consider what tools like Claude Code actually make possible. A manager in a state health and human services department needs to fix a bug in a benefits application—maybe something like the VA submit button Erie describes in her piece, where veterans couldn’t tell their form had gone through. Or she needs to build a new module to comply with a regulation that just changed. Today, she can’t do any of those things without getting budget approval, filing a change request, waiting months (sometimes years!), and paying the vendor an enormous sum. The system is taxpayer-funded and should be owned and operated by the government. In practice, the vendor usually guards access to it. But what if that manager could open an agentic coding tool and make the fix herself?

That may sound dangerous to anyone who’s seen Claude produce code with mistakes or vulnerabilities, but these tools are getting better quickly. At the very least, a small technical staff can do far more than they used to be able to do, putting self-sufficiency within closer reach. The problem of “no one understands the code base except the vendor” should no longer be a problem. Even if you don’t trust Claude to actually push the changes live, it can dramatically collapse the distance between where we are now, with government at the mercy of vendors, and where we could be, with government in control of its own destiny.

Technology is not the barrier to this future—it’s possible today. What makes it unlikely is the norms around who is allowed to touch a system, the way roles have been defined, the procurement rules that created the vendor relationship in the first place, and the contracts that give vendors ongoing control over systems built with public money. These are all aspects of a very broken status quo, one whose time should be up. In the best case, when a philanthropic funder issues a call for proposals encouraging the use of AI, they are creating incentives to overcome those barriers.

Think about how this could play out in the state Medicaid context. States are now implementing work requirements and will each spend tens of millions of dollars on IT systems to do so, with the money flowing to the same small set of vendors, on the same terms, with the same locked-in dependencies. Agentic coding tools could plausibly reduce those costs to near zero for agencies that develop the internal capacity to use them. Remember that these tools are getting better all the time. I find what they can do today remarkable; it’s hard to imagine how powerful they’re going to be if they get even better. If they can be part of a fundamental rebalancing of power between government and the vendor ecosystem, we should use them for that purpose.

AI tools sold by vendors could certainly deepen dependency. The answer is not to avoid AI, though. It’s to be specific about the kinds of AI adoption we’re trying to cultivate and the terms that government should set with its vendors. The kind that builds government’s own capacity is categorically different from the kind that adds another vendor to the stack and doubles down on lock-in.

Philanthropy’s role as catalyst

Lastly, I want to make sure we’re not conflating philanthropic strategy with government policy here. If agencies are requiring AI where its not needed, that’s bad. But philanthropists can’t (and aren’t trying to) dictate what technologies government adopts. The funds they’re dangling are minuscule compared to government IT budgets, but because in government it’s easier to get funding for a multi-hundred million dollar project than for a small test of something unproven, philanthropy is presumably trying to fill a gap.2

When I was starting Code for America, and I got all those questions about why we worked on online services, the other part of my answer was “because that’s what governments are struggling with.” We weren’t philanthropists, but we were raising philanthropic funds to do work in government. To Erie’s point, we would never have said “we only build Ruby on Rails apps” or “we only code in Java,” but we did bring resources to a specific kind of work that was tied to a novel technology because that technology was important and government was bad at leveraging it.

Just because some philanthropists want to encourage experimentation with AI doesn’t mean that even they think that all problems are solved with AI. I certainly don’t. They’re trying to play a unique, catalytic role, shaping the public sector to act more effectively in the public interest. One can believe that the work of debugging a submit button is enormously valuable and high leverage but not believe it’s the kind of thing that philanthropy should fund, for a whole host of reasons. My views on the value of the kind of work Erie describes haven’t changed, but questions of who should do and pay for what and how philanthropy responsibly shapes the public sector require more nuance than just a recognition of what’s valuable. And they require constant updating as progress happens, bottlenecks shift, and circumstances change. Just like the public’s expectations, philanthropic strategy evolves.

Subscribe now

Erie’s piece ends with a question about who benefits from AI adoption in government. It’s the right question. The answer depends on what kinds of tools and how we buy and use them. Tools that paper over and excuse accumulated policy cruft may hurt; those that help simplify it may help. Vendors selling snake oil will hurt; models that empower government teams to own their own destinies should help. AI that digitizes the same broken experience in a new interface will hurt; AI that truly helps meet people’s raised expectations will help. To me, these are not reasons to sit on the sidelines. They are all reasons why philanthropy should try to shape this transition in the public interest.

Share

A gov nerd’s dream collab: Kevin Hawickhorst x Jordan Schneider

Yes, Jordan Schneider had Kevin Hawickhorst on ChinaTalk. To talk about the history of the civil service. And entymologists. And the Bureau of Soils. And most of all, about becoming a serious country again. I’m not even going to summarize it here. Just listen to it. (And if you want more on this, here’s Kevin’s long piece in American Affairs.)

In a much less enlightening episode, but also delightful (at least for me), I joined Jordan as well. I could talk to that guy all day.

Where’s Waldo? (sorry, couldn’t resist): the Intergovernmental Software Collaborative finds a new home — and maybe a new moment

I’ve been loosely tracking the Intergovernmental Software Collaborative since Waldo Jaquith and Robin Carnahan wrote their original paper on how states could pool resources to create the software they need back in 20201. The news last week that the effort is moving from Georgetown University’s Beeck Center to the Council of State Governments caught my eye. It will now be part of the National Center for Interstate Compacts, which already supports more than 250 interstate compacts and recently developed CompactConnect, an open-source data system for state licensing agencies.

Share

The concept of intergovernmental software collaboratives has always seemed like it should work. State officials are often quite frustrated with the choices in the market (or lack thereof) for functions, and complain that vendors abuse their monopoly or near-monopoly positions to sell lousy software at exorbitant prices. And there are also functions that are just so niche that a commercial offering may not make sense. But states have similar needs, and could band together to fund the development of their own software products, and be in control of the development roadmaps. They could get more of what they want, tailored to their needs, for less. In theory.

Examples of where this is working include AASHTOWare for transportation project design and management. It’s billed as “a unique and powerful enterprise software suite designed by transportation professionals for transportation professionals,” as it has users in most states. Or HURREVAC, which helps local emergency managers decide when to issue evacuation orders. Or WinGAP CAMA (the “Win” in its name clues you into its origins in the 1980s), which is actually a county collaborative, allowing 145 of the 159 counties in Georgia to do property tax appraisal for only $2,000 in dues a year (at least back in 2019.) When you’ve got 145 counties, $2,000 each gives you enough a year to keep the software updated, apparently. There are a lot of needs in government like this, and not as many software collaboratives as you’d think.

I believe when I’ve talked to Waldo about this in the past, it’s governance he’s pointed to as the biggest reason we don’t have more of these. Governance of shared codebases can be tricky, but there are best practices that states and other entities can learn from. I mean, look at WinGAP CAMA, which has been making 145 counties happy since 1987. The news that the Intergovernmental Software Collaborative is moving into a new home with so much expertise in shared governance is promising. I’m excited to see what happens.

Subscribe now

But I’m also curious about these collaboratives in the era of AI. The tiny budget of WinGAP CAMA (and probably a few others) notwithstanding, the cost of building and maintaining software has always been a big barrier here. Now, agentic coding tools are collapsing that cost. If states can build or customize at a fraction of previous cost, the case for shared development gets stronger, and the threshold for participation gets lower. The move to the Council of State Governments, with its direct relationships with state leaders and its existing infrastructure for interstate cooperation, sounds like it could be a big unlock, but I’m really curious to imagine what, for example, “software designed by transportation professionals for transportation professionals” looks like when it’s literally designed by transportation professionals, with the software developers needed less and less. With AI, will every transportation department and assessors office just roll their own? How might AI help with the governance issues as well?

The outside-in play for the AI era

April Harding, former director of digital services for the IRS and current We the Doers co-founder, got spicy on LinkedIn last week with a project that reminds of the good old days when Carl Malamud would do crazy shit like putting the SEC EDGAR records online himself, for free. He then essentially harassed the agency to take the website over — which, after many protestations that it would be entirely impossible or hilariously expensive to do so, it did, permanently. The classic outside-in play.

April picked up the Malamud mantel and did in one day what OMB failed to do in 15 years: produce a complete Federal Program Inventory, using ChatGPT. Her diagnosis of why OMB was slow was less about the failure to use currently available technology, and more about a “red ink” leadership culture where executives wait for someone else to produce a first draft before they’ll engage. Nobody owns the blank page. But Harding’s play here follows the Malamud logic: force the draft into existence from outside, and suddenly “prove me wrong” becomes the only thing left to argue about.

AI makes this kind of move vastly cheaper and faster. It could be on the verge of a real renaissance.

Share

We the Doers: doing good things

Speaking of April, (and happy April 1 to all who celebrate!), I don’t know how I missed April and her colleagues’ report How to Achieve Real Government Reform in a Post-DOGE World, which came out in January. It was great to get caught up on the work of this new group, which bills itself as “a civil servant-informed, non-partisan movement to deliver the government the American people deserve.” There were places in the report where I wish they’d taken the ideas further (feedback loops with Congress need to be about more than just getting input on bills up front) and where I have a little skepticism. Will this ultimately be a citizen-led movement? I’m more bought on in it being a civil-servant led movement, which they already seem to be building infrastructure for) but overall, this was a great read. And the organizing here is music to my ears. I’ve long felt that we are vastly underestimating the power of public servants (former and current) working in their personal capacities as a lever for fixing government.

I’m already a big fan of their LinkedIn feed, where I learned about HR 5438, the Incentivize Savings Act, which was apparently introduced in September. It would allow agencies to retain 50 percent of unexpended prior-year funds for modernization and innovation initiatives. We the Doers ran a quick poll, and only six people responded. Let’s get these changemakers more attention so more folks respond to these polls. And I’m very curious about the three people who responded that they wouldn’t relinquish their agency’s ability to ask for a year-on-year funding increase in exchange for being able to retain 51 percent of unexpended prior-year funds. I guess every agency is different. Incentives matter.

I’m eager to hear much more from We the Doers.

Subscribe now

In the mid-1990s, Senator Cohen and Representative Clinger asked the right questions about the federal government’s technology problem. The Brooks Act and other legislation had established technology in government as a primarily a purchasing problem, and agencies had no one whose job it was to think about technology strategically. Clinger-Cohen established a CIO Council and required federal agencies to develop coherent IT architectures. It told agencies to measure outcomes, empower CIOs, and treat technology as a strategic asset. These were very reasonable remedies.

Share

But the law was written in 1995 and passed in 1996. The World Wide Web had been invented six years earlier. Amazon launched the same year Clinger-Cohen was written. Google was three years away. The consumer internet was about to transform society more profoundly than anyone in the halls of Congress fully grasped. After decades of failing at technology, Clinger-Cohen tried to catch government up, but just as it brought some reasonable answers, the questions changed.

That same year, Congress passed the Paperwork Reduction Act of 1995, a major revision of the 1980 original. The original PRA had created the basic framework, OMB oversight of paperwork burden. The 1995 rewrite locked in the specific procedural machinery: the requirement that agencies get OMB approval before creating any form, website, survey, or electronic submission, the 60-day and 30-day public comment periods, the three-year renewal cycles. It was a thoughtful, bipartisan effort to reduce the burden of government information collection on the public, and like Clinger-Cohen, it came at an awkward time.

The internet was about to turn “government information collection” from paper forms mailed to your house into something vastly more dynamic, interactive, and frequent. The PRA’s procedural apparatus was designed for none of it. As I’ve argued at length, the PRA has never done what it claimed to do. It has not reduced burden on the public. But it has added enormous compliance costs inside agencies, delayed critical changes, and made it nearly impossible for agencies to test and learn their way to better services. The specific processes that Congress codified in 1995 locked agencies into processes that were almost immediately outdated.

Subscribe now

So here we are in our own version of 1995. The lesson of both Clinger-Cohen and the PRA isn't just that the timing was bad. It's that no single law could have done what the moment required: not only harness technology in the service of existing government, but anticipate the changing needs and expectations the country would face and build a government capable of meeting them. We are just starting to understand how government might use AI. We don’t really understand how AI will change the work government needs to do. So we risk boldly dragging government into the mid-to-late 2010s and then letting it stagnate for another generation.

Meanwhile, the list of laws governing government technology that are overdue for replacement is long. Beyond the PRA, you have the Privacy Act of 1974, designed to protect citizens from government overreach in the era of mainframe databases, now preventing agencies from pre-filling information they already have, forcing people to provide the same information repeatedly to different parts of the same government. FISMA offers a reasonable menu of security controls, but in practice agencies implement nearly all three hundred of them regardless of relevance, adding months or years to timelines while actually making systems less secure by crowding out time for testing. All these and others represent a particular era’s best guess at how to build and buy technology. All of them have been overtaken by events.

These laws need to change. This is not a partisan position. Anyone who has complained about government’s inability to deliver (who hasn’t?) should be supporting bipartisan work to modernize them. AI’s evolution and adoption are happening fast. The distance between what technology makes possible and what our legal framework contemplates is about to grow massively, and it will keep growing. We must act, but we need to avoid the mistakes of 1995. And I think we can.

I’ll be writing over the coming months about how I and others think about getting it right this time. The short version is that we need to stop specifying processes that sound reasonable but have not been tested in the real world, do not have built-in mechanisms for revision and adaptation, and will not survive the transformation that’s already underway. And we need to acknowledge that even the best-designed legislation fails when it’s aimed at agencies that lack the capacity to deliver on it. In addition to purpose-fit systems, government needs the right people, focused on the right work, and incentivized towards outcomes. It needs oversight frameworks that support the experimentation that learning requires. These ideas deserve more than a paragraph, so stay tuned.

Share

For now, I want to be clear that I’m not saying we should wait until we understand AI before we reform. We can’t afford to wait. The PRA is still haunting service delivery. The Privacy Act of 1974 is still preventing sensible data sharing. Delay is its own form of failure. What I’m saying is that the pattern — write reasonable-sounding laws, lock in the mechanisms, watch them calcify — is not inevitable. We know enough now to choose differently.

Subscribe now

Subscribe now

An Opening Move on Civil Service Reform

A Civil Service for the Mission, Margaret Mullins, Vanderbilt Policy Accelerator

This is a truly comprehensive proposal for genuinely restructuring — not just tinkering with — the federal civil service, and it comes with actual draft legislation. Whether or not you agree with every element of Mullins’s specific prescription, she’s doing something important: making the case, rigorously and with receipts, that we have to do something, and that every other reform we can imagine depends on it. You cannot fix government delivery if you can’t hire people capable of delivering. You cannot modernize agencies if the people inside them are trapped in classification and compensation structures designed for a government that hasn’t existed since 1949.

The historical diagnosis is the strongest part of the paper, and worth reading even if you end up disagreeing with the proposed remedy. Mullins traces how nearly every reform effort since the Pendleton Act has consolidated power in the executive while leaving the actual machinery — hiring, classification, compensation — untouched. This is not an accident. Controlling who gets fired and who answers to whom is politically useful in a way that fixing job classification standards simply isn’t. The result is a civil service that is increasingly responsive to whoever holds executive power, but no more capable of actually doing the work of government. Worth noting that this dynamic cuts in a complicated direction: the civil service has become more responsive to executive power at the top — through the Senior Executive Service, Schedule P/C (see below), and related tools — while remaining famously resistant to it at the level of individual performance management, where removing an underperforming employee is still extraordinarily difficult. So we’ve managed to build a system that is both too politicized and too rigid at the same time, which is quite an achievement. The Civil Service Reform Act of 1978, often cited as the last serious attempt at comprehensive reform, was, as she shows, comprehensive in name only. The Reagan and Clinton administrations then made it worse by decentralizing responsibility without giving agencies the capacity or guidance to handle it.

Share

There will be other proposals, and there will be productive dialogue among them. (My Niskanen colleague Gabe Menchaca is working on a massive, much-anticipated tome, which will agree with Margaret on the disorder but disagree on remedies.) Civil service reform is one of those rare areas where the dysfunction is so widely acknowledged — across party lines, across administrations, across ideological traditions — that a bipartisan path forward is possible. It will be harder to build the political will to do the boring, unglamorous work that actual reform requires than to diagnose where we went wrong. But there are a lot of us up for that task right now.

Three Takes on OPM’s New Schedule Policy/Career Role

• Loren DeJonge Schulman, Everything You Need to Know (and Ask!) About OPM’s New Schedule Policy/Career Role: Oversight Resource for OPM’s Schedule Policy/Career Rule, Federation of American Scientists

• Don Moynihan, Trump’s Schedule F Rule Finalized, Can We Still Govern?

• Ronald Sanders, The ‘new’ Policy/Career Schedule does NOT (necessarily) politicize the federal workforce, Government Executive

The newly finalized Schedule Policy/Career rule, the renamed and revised version of Schedule F, moves a broad category of “policy-influencing” career federal employees out of the competitive service, stripping them of Civil Service Reform Act due process protections and MSPB appeal rights — making them, in effect, at-will. As evidenced above, there’s a wide and somewhat surprising range of views here.

Schulman’s FAS piece is the overview you want if you’re trying to understand what the rule actually does — it’s a much-needed a practical resource for congressional staff and the rest of us, structured around what to watch, what to ask, and where agency discretion is likely to determine outcomes. She is clear-eyed that the rule’s real impact depends almost entirely on how agencies apply it.

Moynihan is considerably more alarmed, documenting the ways the rule misrepresents the evidence, dismisses the overwhelming opposition in public comments, and ignores the politicized firings and reprisals of the past year as if they simply hadn’t happened. As usual, Don has a point.

Sanders, a longtime SES member who actually resigned from a Trump 1.0 appointment over the original Schedule F, lands somewhere unexpected: he thinks this version is meaningfully different. OPM Director Scott Kupor’s implementing guidance — which provides that employees need only follow lawful orders “faithfully and to the best of their ability,” not pledge personal loyalty — is, in Sanders’s view, a real guardrail, and one that Sanders says would have been enough to keep him from resigning the first time around. Whether it holds is essentially a question of whether OPM polices its own guidance. Which brings us back to Schulman.

Where one might come out on this depends not just on your politics but also on whether you expect to trust future administrations that will be implementing these guidelines past the current moment. If you think you will, it’s easy to read them as written in good faith to make the civil service more responsive to political will — which benefit whichever party or faction is power, of course. If you think you won’t, then we’re trapped in the same cycle of hobbling government to constrain our opponents that got us here in the first place. I have no crystal ball. I wish I did.

Subscribe now

Tech Force Now in Force

Speaking of cutting both ways, Tech Force has people now, and there’s been concern about an ethics rule that allows private sector technologists to serve while remaining vested at their current companies. This came up during the Obama years too. There was not a snowball’s chance in hell we could have gotten that ruling, or that anyone would have even tried, but it’s worth saying that any administration trying to recruit top tech talent thinks about how to make it easy for them. Back then, plenty of folks were happy to leave their companies, knowing they could easily go back, so I never wanted to pursue it, especially because, yes, there are real conflict of interest issues there. But when the shit hits the fan, any administration wants the people who can fix it, ASAP. Remember healthcare.gov? And AI is definitely hitting the fan.

While we’re on the topic, many people assume I am hostile to Tech Force because it (sort of) replaces USDS, which I helped found. In fact, I wish them great success. Abigail Haddad has a clear-eyed view on whether technologists should sign up, including why this work has always been hard, under any administration. She comes down as a maybe, which I think is fair, but I hope lots of people find opportunities they feel comfortable with to serve the American people and be a part of what should be a dramatic transformation of government tech in an era of AI. To quote a former 18Fer:

“Thinking people only deserve a better government during an administration you support is the height of cynicism in a truly golden age of cynicism.”

Technically compliant, practically failing

GovIntegrity / PIA, Necessary. Not Sufficient. (Note, I wish we’d get an actual author on these posts, and GovIntegrity needs a proper about page on their substack!)

Moving on from civil service, to the extent that one can in the sense that it underpins every other reform….

The Program Integrity Alliance has a substack, and it’s a hidden gem. I thought their explainer of what happened with fraud in Minnesota was the best I read (spoiler: it was not a failure to detect fraud, but rather the lack of “clear, court-defensible authority to temporarily pause payments when credible fraud risk is documented.”) But their piece on income verification is great (and rhymes with hometown hero Luke Farrell’s exceptional take on the same), their observations about reliance on paper is critical in the age of AI, which is excellent at generating documents, and there’s a lot more there. I used to have a bit of an allergic reaction to talk of program integrity, as it certainly can get weaponized in the service of denying legitimate claimants much-needed benefits, but I’ve come to believe that program integrity done right is absolutely critical to the reform agenda, and that the observations this community shares are deeply resonant with those of the state capacity community. They are constantly pointing out how systems can “technically comply with process while remaining structurally incapable of distinguishing paper compliance from real service delivery.” Yes, exactly.

Share

Their most recent post is on OMB’s recently revised Circular A-123, which governs how agencies assess and manage operational and financial risks. The update explicitly adds fraud prevention to the list of things agencies must take seriously as a management responsibility — not just a law enforcement problem to be handled after the money is gone. That framing shift matters, and the author credits it as a meaningful step. But the piece is really about all the ways the step falls short: the circular runs through CFO offices that have limited authority over the program and IT systems where fraud actually happens; outdated federal privacy statutes prevent the cross-program data matching that would let agencies see organized fraud rings working multiple programs simultaneously; and nothing in the incentive structure changes. I would quibble with some of PIA’s solutions; to me, mandating real-time transaction screening tools that could catch fraud before payment is too prescriptive, in the sense that those tools will evolve and we want agencies to keep up, not comply with guidance that’s stuck in a moment in time. But the statutory constraints point hits close to home for me. We keep bumping into the same pattern: well-intentioned executive branch guidance that can’t actually solve the problem because the underlying legal framework hasn’t been touched. Fraud prevention joins a long list of things that require Congress to act, not just agencies to try harder.

PIA is clearly an important voice; I’ll be following what they write.

That’s it for today. See you next week…maybe. Not making any promises!

Subscribe now

Move Fast. Fix Things.

A speech by Darren Jones, Chief Secretary to the Prime Minister, UK Cabinet Office — Jan. 20, 2026

Like my collaborator Andrew Greenway’s recent screed (and where I found this), this is worth reading from across the Atlantic. Darren Jones, the UK’s Chief Secretary to the Prime Minister, gave this speech in January laying out a plan to rewire Whitehall — and the diagnosis will feel familiar. Years of bureaucratic layering have produced a civil service that spends more time checking than doing, more time on internal policy papers than on delivering services to citizens. Jones’s proposed remedies are concrete: a new framework rolling out in April that cuts approval processes (one HMRC pilot reduced 40 required approvals to 2, saving two to three months of delivery time), new hiring criteria for senior civil servants that weight frontline delivery experience over policy-writing ability, and performance management with actual teeth. The title — a deliberate riff on Silicon Valley’s “move fast and break things” — makes the point that speed and care aren’t opposites. The UK hasn’t necessarily figured this out any more than we have, but they are naming the same problems out loud and commiting to specific structural changes.

A question for the comments: If the equivalent of this speech were given in the US, who would give it?

Subscribe now

Plain Job Titles Please

State Capacitance / Kevin Hawickhorst, Foundation for American Innovation — Feb. 5, 2026

My favorite gimmick of the year so far is Kevin Hawickhorst’s Days Since Last ‘IT Specialist’ Posting tracker.

If you want to scare qualified software engineers off of applying for your open role (or ensure they never see it), make sure you call the position an “IT Specialist.” You can post design, user research, and product management roles as IT Specialists too. (While you’re at it, call practically everything else a “Program Analyst.”) It’s a fantastic way to make some of the most meaningful and important jobs in the country sound like an opportunity to change the toner cartridge in bleak offices around the federal government.

IT Specialist comes from an OPM publication issued in May 2001 (that’s 25 years ago, but who’s counting) entitled “Job Family Standard for Administrative Work in the Information Technology Group, 2200” which establishes the authorized official position titles for IT Management. It was the standard job title until September 2025, when the Office of Personnel Management (OPM) issued guidance in response to Executive Order 14170 instructing agencies to stop using “generic or jargon-laden” job titles and to replace them with titles that are “descriptive, organizational, or functional in nature.” Many have. But as the Foundation for American Innovation’s tracker shows, some agencies are finding it hard to move on.

FWIW, New York has it worse. Their tech jobs are all called “Computer Associates.”

When FAFSA Broke, They Called This Guy

Statecraft / Santi Ruiz — 2026

Jeremy Singer, President of the College Board, spent six months inside the Department of Education helping rescue the botched FAFSA rollout. Santi does his usual expert job of drawing out the story. The chaos was caused by a convergence of the familiar culprits: legislative drafting that was precise in intent but catastrophic in execution, political leadership with its eye on a different prize (loan forgiveness), and the by-now-ritual inability to communicate honestly with the public when things go wrong. Listen in particular for Singer’s observation about why the lessons of healthcare.gov didn’t seem to make their way to the Department of Education, or throughout Democratic leadership.

One Neat Trick for Buying Software That Isn’t Trash

StateScoop / Waldo Jaquith — Feb. 2026

Spoiler alert: the trick is just hiring people who actually understand software. Waldo Jaquith — government delivery manager at U.S. Digital Response and a veteran of this fight — makes the deceptively simple argument that agencies can’t buy good technology if they don’t employ people who understand it. Without that internal expertise, you can’t write a sensible RFP, you can’t evaluate vendors, and you end up getting precisely what you asked for — which is to say, trash. The logic is the same one we’ve been making about state capacity for years: you can’t outsource your way to competence. Given that we just watched the federal government shed hundreds of its most capable technologists, this argument feels newly urgent.

What Breaks Us

Bryce Fountain, LinkedIn — 2026

Saw this randomly on LinkedIn. Bryce Fountain served eight years in the Air Force. He’s done. And the thing that broke him wasn’t the deployments or the danger — it was six months waiting to be reimbursed $5,000 in legitimate expenses because the Defense Travel System, which apparently runs on logic from the 1990s, couldn’t process a trip that got changed and then canceled early. Meanwhile, a contractor traveling alongside him emailed receipts to HR and had money in his account the next day. It’s easy to talk about “procurement reform” and “legacy IT modernization” in the abstract and lose sight of what bad government software actually costs — not just in dollars, but in the people who walk away because they’re done being treated, as Bryce puts it, “like second-class citizens.” A lot of very smart people have tried to fix the DTS. Someone tell me in the comments why we’re still stuck here.

Blue Cities and States Are in Trouble. Democrats Need to Change How They Run Them.

New York Times Opinion / Nicholas Bagley & Robert Gordon — Feb. 23, 2026

The authors of this op-ed are both veterans of Democratic administrations, and both colleagues of mine — Robert is at Recoding America (though he is writing in his personal capacity here) and Nick and I are both senior fellows at the Niskanen Center. They are saying something Democrats don’t often say out loud: the party’s relationship with public sector unions is making it harder to govern, and something has to give. Their argument isn’t anti-union in any simple sense; it’s that the current arrangement lets unions prioritize the interests of their existing members over the quality of services that the public — and especially lower-income communities — depend on. They’re calling for a new bargain: one that protects workers’ voices and livelihoods but puts public services first. The piece may be uncomfortable for some, but it’s pointing at a structural problem that’s been quietly obvious to anyone who’s tried to reform a public system from the inside.

Subscribe now

Is a New Teacher Better Off in Mississippi Than in New York?

Slow Boring / Matthew Yglesias — March 5, 2026

Almost a companion to the Bagley/Gordon piece, Yglesias starts from a fact that should alarm anyone who cares about public investment: we are living through a surge in anti-tax politics, and it’s not just coming from the right. Across the political spectrum, people are losing faith in the proposition that paying more taxes produces better outcomes they can actually see in their lives. Teacher compensation is his case study. New York spends vastly more per pupil than Mississippi or Louisiana — and is getting roughly comparable outcomes. Mississippi and Louisiana, meanwhile, have been posting some of the biggest educational gains in the country, largely through curriculum reform that costs relatively little. So are high spending and good outcomes as correlated as the left would like to believe? Yglesias’s argument is that if progressives want to make the case for public investment — in schools, in services, in anything — they first have to be able to show that existing money is being well spent. Meanwhile, in California, newcomer to the governor’s race Matt Mahan seems to be the only candidate recognizing this.

I’ll probably be saying this a lot this year: capacity isn’t the number of people or the size of the budget, it’s the demonstrated ability to achieve policy goals.

Bureaucracy As Social Hope: An Argument for Renewing the Administrative State

Federation of American Scientists / Hannah Safford, Loren DeJonge Schulman, Craig Segall, and others — Feb. 12, 2026

FAS just launched a new Center for Regulatory Ingenuity, and this essay collection is its opening argument. It starts from the observation that every major political faction has arrived at the same conclusion — that the administrative state doesn’t work — but they’re telling very different stories about why, and proposing wildly different remedies. The authors’ counter is that the government we have was largely designed to move slowly and deliberatively, to resist capture by any particular interest. That design produced real successes. The problem is that the tools that got us here aren’t the tools that will get us to where we need to go — and we are now, as they put it, “starting to hit real limits.”

The collection uses climate policy as its lens, but the insights travel well. The essays, which will be published over the course of the coming weeks, range from Jordan Diamond and the Environmental Law Institute on the limits of 1970s pollution statutes for driving a 21st-century energy transition (out now), to Loren DeJonge Schulman and Shaibya Dalal on public participation as a governance asset rather than a compliance burden. Hannah Safford and I are working on our contribution to this series, on a critique of government’s obsession with planning, or rather, the tendency of government to treat the plan as the destination rather than the compass. (Eisenhower said it well: “In preparing for battle I have always found that plans are useless, but planning is indispensable.”) The collection’s conclusion is bracing: “taking a chainsaw to government leaves our whole nation bleeding” — but neither is defending the status quo a plan.

A Premortem on OPM’s HR 2.0 Initiative

Niskanen Center / Steve Krauss, Gabe Menchaca, Peter Bonner — March 2026

Rather than waiting for the inevitable GAO report or congressional hearing that will someday explain why the federal government’s effort to consolidate its sprawling HR systems fell short, the authors — all former senior officials at OPM and OMB — have done a premortem: imagining failure now, in order to avoid it. The underlying problem is real: the federal government runs dozens of incompatible HR systems, hundreds of interfaces, and countless manual workarounds that are costly, brittle, and increasingly unable to support how government actually needs to recruit and manage people. The Trump administration’s HR 2.0 initiative aims to fix this. Maybe it will! But the authors want to make sure we go in clear-eyed about how these things tend to go wrong.

Share

The Melting Pod: Biden and the Border

I’ve just started this brand new podcast from alums of the Biden admin, so caveats apply, but it asks a question that always fascinates me: How can many caring, smart, well-meaning people thoroughly bonk something? I’ve been warned there’s no facile answer here, which generally checks out, and I appreciate the honest tone of the pod so far. I’ll be listening for state capacity lessons that might apply under any administration.

That’s it for the first roundup. If you like it, maybe I’ll make it a habit!

An experiment out of the California legislature may provide an illustrative example. On Thursday, California Speaker of the Assembly Robert Rivas announced an intentional, structured process for evaluating whether the laws lawmakers pass actually do what they’re supposed to do. The program is called Outcomes Reviews, and it’s launching as a modest pilot in the state’s lower house, with the intention to look back at roughly 8-15 laws in 2026. By starting small, this initiative is hoping to test and learn how to build the muscle of test-and-learn, so to speak

The problem the Speaker’s office hopes to address is that California passes a lot of bills, but those bills aren’t necessarily solving the problems their authors are targeting. Between 2016 and 2022, for example, more than 100 bills to stimulate housing production were signed into California law, but even today permit numbers in the state have barely budged. This is the problem that sits at the heart of the Recoding America Fund’s work: government’s poor track record of delivering on policy intent.

Subscribe now

Outcomes Reviews are an experiment in closing that gap. They’re essentially a process that starts with a legislator announcing that they’re going to review the outcomes of a bill they previously championed. The member coordinates with the relevant policy committee and identifies stakeholders who’ll be part of the work. Then, over the course of a legislative year, the legislator and their staff, working with committee experts and direct stakeholders, hold a series of roundtables and hearings. They listen to the people actually affected by the law. They dig into what’s working, what’s not, where implementation has stalled, where unintended consequences have emerged. Finally, they announce what they’ve learned and what they’re doing about it. This might be a fix-it bill, an executive action request, or public coverage celebrating the law’s success—whatever the evidence suggests is needed. The work helps close the feedback loop between policy intent and real world outcomes.

Since Speaker Rivas took the gavel, he’s been focused on shifting the focus of his chamber from the volume of legislation to meaningful outcomes that matter to Californians, starting with reducing the bill cap for each individual member from 50 to 35. But legislators introduce bills in part because doing so gets them attention. A known, predictable cadence of activity is built into the calendar. Press, stakeholders, and advocates all react to the steps in the process of how a bill becomes law (and many of those steps occur even if it doesn’t become law.) Attention gets you reelected. Passing a bill is the most well-worn, reliable path for members to that attention.

Share

Outcomes Reviews are designed to be that alternative path. By offering the same sequence of public moments—announcement, engagement with stakeholders and advocates, deliverables with media attention—they provide a second legislative rhythm—one focused not on passing new laws, but on making existing ones work. If it sounds like oversight, it is, but less reactive and less outrage-driven. The structure is designed for more sustained, thoughtful inquiry than is common in the usual hearings in response to high-profile catastrophes, which can often devolve into a blame game.

The announcement from Rivas’s office provides a few examples, starting with AB 2011, authored by Assemblymember Buffy Wicks and known as the Middle Class Housing Act. It was enacted in 2022 to make it easier to build affordable and mixed-income housing projects in cities and metro areas where shops or offices are already allowed. Wicks will use an Outcomes Review to dig into whether the law is actually achieving its purpose. She’ll hold hearings with developers, city officials, and community members. She’ll examine whether the reforms are sufficient, whether there are unintended barriers, whether the problem the law was meant to solve is actually being solved. And then she’ll announce what she’s learned and what comes next. That announcement will hopefully get media coverage. Advocates should engage with it. Her constituents should see her following through on the problem she seeks to solve, not moving on to bill number 36.

Assemblymember Jacqui Irwin is set to dig into the results of Assembly Bill 488, enacted in 2021 to ensure charitable donations have their intended impact. Irwin will use an Outcomes Review to examine how the law is functioning and, specifically, how it’s serving victims of disasters like the Los Angeles firestorms. This isn’t a new bill, but it comes with the same political infrastructure—the press conference, the stakeholder engagement, the announcement of solutions—that makes introducing a bill attractive.

AB 457 was enacted in 2025 with the goal of building more affordable farmworker housing within 15 miles of farm or grazing land in the Central Valley. Its sponsor, Assemblymember Esmeralda Soria, had this to say about why she’s piloting an OR.

“Our constituents deserve to know whether the laws we pass are being implemented and are effective. They should know how these laws are working or not working and impacting their communities. The Outcomes Review tool will ensure transparency, oversight, and accountability at a time our constituents look to us for solutions. With these new outcome review tools, our constituents will know whether more affordable homes are being built for California farmworkers, as AB 457 clearly lays out.”

Appropriately, the Speaker’s office is treating this new initiative as an experiment. There are many questions they may be hoping to answer: Do members have the bandwidth and expertise to do this well? Can legislators actually benefit—politically and reputationally—from doing this backward-looking work, or will the culture of “move on to the next bill” persist? Do advocates and the press actually engage with Outcomes Reviews the way they do with new bills? Do the reviews actually identify problems? Do the solutions get adopted? Do laws improve as a result?

Subscribe now

My big question is whether this mechanism allows for the deeper kind of inquiry that gets at the underlying issues in bureaucracies. Most oversight increases risk aversion because it tends to point fingers at administrative agencies for implementation failures and it tends to happen when something has gone wrong in a highly visible way, making outrage the natural response. Outrage discourages public servants from trying new things; you may get yelled at but you can’t get in significant trouble if you’ve played by the book. A deeper dive at a less heated moment and over a sustained period has a greater chance of surfacing the real barriers these public servants face. If legislators really listen, they’ll hear the things that they need to do to speed implementation in agencies, things like making it easier to hire the right people, to reduce unnecessary procedural burdens, and to make it possible to build and buy the systems that administer programs. That’s critical work our legislators should be focusing on.

Kudos to Speaker Rivas for this bold experiment. I look forward to seeing what happens as the California Assembly tests the program, documents what happens, and learns what’s hard and what works. Perhaps there will be tweaks and adjustments, and hopefully it will eventually merit a broader roll out to all 80 Assembly members. A future in which Outcomes Reviews are a large part of what the press, advocates, and constituents expect from their legislators — and we meaningfully close the loop between lawmaking and implementation – is a better future.

But California doesn’t need to be the only test bed. Other states can, and should, try their own versions of this. Not every state is the bill mill California has become, but every state, and of course our federal government, needs to care less about passing laws and more about making good on the promises they represent.

Share

Subscribe now

He goes on to say:

American progressives want a Swedish-style welfare state without doing any of the hard work that is involved in creating a state apparatus capable of delivering a Swedish welfare state. There are thousands of examples of this, but one that I found particularly striking was Elizabeth Warren’s call for the introduction of a wealth tax. This is a typical “all-pudding no-meat” position. There are various objections to wealth taxes, but one of the most significant is that imposing them requires the creation of an entirely new administrative system, since taxpayers are currently only required to report their income to the government, not their wealth. You can’t just add a few extra lines to the income tax form. The problem is that in 2020, when Warren was campaigning on this, the U.S. federal government barely had the capacity to collect an income tax.

Heath then proceeds to tell a personal IRS horror story, an 18-month long saga of threatening letters directing him to talk to an agent despite there being essentially no agents to talk to. I won’t repeat it here, but it’s the kind of story that reminds me that the people complaining that government efficiency is a terrible goal probably haven’t been tortured by an agency for 18 months. (If you’re wondering how an agency gets to a state where it treats people like this, despite many well-intentioned efforts to improve it, I recommend Sam Corcos’s interview on Modern Wisdom, in which he shares that the IRS has 108 competing sources of truth and 50,000 active fax lines, not because people want to fax them, but because they still require that certain information be transmitted that way.) Heath wraps up with this observation, with which I wholeheartedly agree:

The larger point is that the IRS is barely capable of administering an income tax. The idea that they could administer a wealth tax is like something out of science fiction.

Zohram Mamdani will become Mayor of New York City on January 1, and the question on my mind is whether his administration will indulge in its own fictions or get to the hard, unsexy work ahead. I hope they eat their vegetables first. He’s promised free child care, for instance, but faces challenges not only funding it, but implementing it. New Mexico, which has a quarter of the population of New York City, is one to watch here. Even before its new policy came into effect, the state had one of the most comprehensive child care policies in the country, but less than one third of eligible families took advantage of it. The Atlantic’s analysis of the challenges New Mexico faces look a lot like the elements of the operating model we describe as the work of the Recoding America Fund: for government to be capable of achieving the policy goals its sets, it needs the right people, focused on the right substantive work, with purpose-fit systems, and test-and-learn frameworks. For New Mexico – and New York City — to deliver, it needs to hire a lot of people well and fast, streamline and right-size licensing and certification procedures, manage these processes through appropriate technology, and adjust along the way as the state learns from what’s working and what isn’t.

For team Mamdani, their first challenge is already clear: hiring. Apparently they got 25,000 job applications in the first 24 hours.1 A lot of people want to work for the Mamdani administration. This is great! They will need a lot of people! The problem is that hiring in New York City government is subject to possibly the most insane rules in the country, as my colleagues Robert Gordon and Gabe Paley have just described for Vital City. And I say that as someone who didn’t think anything could be worse than the federal hiring rules.

Say you work in New York City government and need to hire someone. (Note: I researched this several years ago when I was writing my book, so if it’s changed since then, forgive me, and I will update this if I get corrections.) Your starting place is not an accurate description of the work that needs doing, but rather a desperate search through the job classifications system, where about 3000 different positions, a mix of totally generic, oddly specific, and generally outdated, are described.

This doesn’t sound bad, (and indeed, the truly crazy parts are still to come) but let me explain what often happens when you have job classifications that are arbitrary and out of date. I learned these pitfalls from federal job classification system, when my colleague Marina Nitze tried to hire a web designer at the VA in 2014. There was no job classification for web designers, and creating new classifications takes about a decade. (A recent example is the addition of data scientist, which throughout the 2010s was called the “hottest job of the century,” and highly relevant to many public sector functions. OPM finalized the new Data Science Series, 1560 job classification in August 2021, just in time for data science to be taken over by AI as the hottest job.) Without a relevant classification, the VA personnel team borrowed the criteria from a different tech-related job description, and with it, the qualifications for that job. Though we know of a dozen highly qualified and in fact hugely talented people who applied (because we encouraged them to!), when Marina got her “cert” (a list of the candidates that HR has deemed qualified, from which the hiring manager is supposed to choose), none of them were on the cert. Those who were held various certifications for Microsoft, Cisco, and other vendors’ products and platforms appropriate to someone doing network services or systems administration. None of them had any experience whatsoever in web design.

Share

Marina would have had the exact same problem in New York City. As my colleagues point out:

Among 3,000-plus standard titles in city government today, not one is for a technology job recognizable to ordinary technologists — no “software engineer,” no “web designer.”

The city is, however, presently offering exams for three kinds of “computer associate,” covering everything from allocating desktop computers to coding.To learn more about the “computer associate” role, “YOU ARE RESPONSIBLE FOR READING THIS ENTIRE NOTICE BEFORE YOU SUBMIT YOUR APPLICATION.” In that seven-page PDF, most of which is consumed with describing educational qualifications and relevant certifications, here is the entire description of what the job is:

Computer Associates (Operations), under general supervision, with very considerable latitude for independent initiative and judgment: supervise the activities of subordinates in one or more computer operations units of considerable size; or serve as a technical resource person in the performance of networked, multi-tiered, or mainframe computer operations; or perform as a technical resource person in the diagnosis of and, when feasible, the correction of telecommunications hardware problems in order to maintain efficient functioning of telecommunication operations and to minimize downtime in the case of system failure. All Computer Associates (Operations) perform related work.

Telecommunications hardware! Exactly.

My colleagues and I mention technology jobs because they’re good examples of how outdated the system is, but the point holds for jobs of all flavors. And the problems in New York go much deeper than a need to update job descriptions.

In New York, a hiring manager can’t just pick the closest job title and description and hire someone using it. And an applicant can’t just apply for an open job that interests them. A central hiring authority periodically conducts exams for some of these 3000 different positions, and applicants interested in that position must take the exam when it is offered if they want to be eligible. For example, one of the positions is Administrative Business Promotions Coordinator. If, as an applicant, you think you may one day want to be an Administrative Business Promotions Coordinator in any agency or department anywhere in the City or State of New York, you must watch for when that exam is administered and spend $85 of your own money to take the exam. If your life goal is to work on transportation, or the environment, or public health, and you don’t know exactly what job title you would need to hold, you’re out of luck, because you can’t apply for a specific agency like the Department of Transportation, only for these generic and often murky job descriptions that may (or may not) qualify you for a job that comes up under that position description anywhere in government in the future.

But as an applicant, you really have no idea what title the job you may want in the future may be given, and that’s because the hiring managers and their HR partners must play an absurd game when they have a position that needs filling. Let’s say you work in the Department of Transportation and you’ve decided you need an Administrative Business Promotions Coordinator. (If you don’t quite know what that is, no worries. We’re in the same boat.) If an exam has been conducted for Administrative Business Promotions Coordinators in the last five years, as a hiring manager, you can only interview the three people who took that exam and scored the highest. You must hire one of them, no matter if they are a fit for the job you are trying to fill. None the people out there who are knowledgeable and passionate about public transportation and actively looking for a job are eligible, unless they happened to have seen into the future a few years ago, known that a job for Administrative Business Promotions Coordinator would come up in the Department of Transportation, found their way to the obscure web page that would have told them this exam was happening, shown up in downtown Manhattan on a Saturday, waited for several hours to take this exam, and rated among the top three on what might have been a test of attention to detail or might have simply been a questionnaire asking them to rate themselves in various competencies. Because the pool of people who did take the test is often skewed towards those with no significant interest in the domain but with a strong interest in a steady job from which they almost can’t be fired, and the rules around hiring off the list are so stringent, hiring managers have developed ways to avoid hiring off the list at all.

What do they do instead? They look around for a job title for which no test has been administered in the last five years. There are often many of them, but the job title they pick must also be one the agency has the authority to hire, as not all job titles are allocated to all agencies. If they can find a job title that’s somewhat defensible as a descriptor of the work that needs to be done, for which a competitive exam hasn’t been administered in the last five years, and for which the agency has the authority to hire, that’s the job title they will pick because then they can run their own competitive process and select someone better suited to the job. It can take weeks of hunting around various job titles to find one that fits those criteria before the process can even begin. One public servant told me that job title hunting is the main skill HR professionals in NYC government are valued for. There’s no time for developing candidate pools or creating effective selection processes. Almost all of the capacity of HR professionals in NY city and state government is eaten up by a game they must play if they want even a shot at hiring someone fit for the job. Ironically, the game tries to outwit a system that’s supposed to ensure merit hiring.

We have these supposedly merit-based hiring systems because in the mid- to late-19th century, positions in government were filled through patronage, or what’s known as the spoils system. After a party won an election, the leaders gave out jobs in the new government to friends, family, supporters, and others to whom they owed favors. In 1881, a desperate crackpot named Charles Guiteau believed that the newly elected President James Garfield owed him a diplomatic post as a reward for an incoherent and ineffective speech he had printed and distributed in support of Garfield’s candidacy. (The speech had previously been written in support of former President Ulysses S. Grant, and when Garfield beat Grant out for the nomination, Guiteau awkwardly and incompletely attempted a 19th century version of search and replace of Grant for Garfield. The result was not good.) When Garfield failed to grant him the post, or to pay him any attention whatsoever, Guiteau shot him at point blank range in a waiting room at the train station, only four months into his presidency. Garfield died 79 days later.

The assassination brought attention to the problems of the spoils system, and created the conditions for the passage of the Pendleton Act of 1883, which established “fair and open competition for federal jobs, admission to the competitive service only on the basis of neutral examination, and protection of those in the service from political influence and coercion.” Though it covered only a small fraction of government employees, it was the first time federal employees were both hired for merit and protected by law from being fired for political reasons, and it served as the basis for future legislation that both expanded the number of employees covered and strengthened the protections. Among these was the Classification Act in 1923 which “established in law the principle of nationally uniform compensation levels, providing for the standard classification of duties and responsibilities by occupations and positions with salary levels assigned to the resulting positions.”

These “duties and responsibilities by occupations” got an update with the Classification Act of 1949, which directed the Office of Personnel Management to prepare standards for agencies to use in placing positions in their proper classes and grades, resulting in the creation of the General Schedule, which is still in use today for most federal civilian employees. Certainly changes have been made to the GS since then, but mostly to make them more complicated and ossified.

New York City’s hiring rules are different, but they derive from New York State law that was inspired by the same events that led to the Pendleton Act in the same era, and extended to every city, town and village in New York. Robert Moses spent the early part of his career attempting to reform New York’s civil service system. His biographer Robert Caro attributes his Moses’s famous thirst for power to his early failure to effect change in the civil service system. To build his empire, he created various authorities and public benefit corporations like the Triborough Bridge and Tunnel Authority (which became Moses’s primary power base) and the Jones Beach State Park Authority that were exempt from the civil service rules he abhorred. The workarounds got him what he wanted, but left the system in place for the rest of the city. The resulting power imbalance between team Moses and the rest of the bureaucracy drives the story we all know, with its lessons for good and ill.

Subscribe now

That’s not a story Mamdani probably derives inspiration from. But the New Yorkers who voted for Mamdani were inspired by his big, bold visions. Big, bold visions rely on strong foundations. The civil service system is the bedrock of that foundation. If Mamdani wants to achieve a fraction of what he’s promised voters, he should start with Moses’ unfinished business.

Share

UPDATE: For concrete, detailed recommendations, the Mamdani team should look to Rebecca Heywood’s excellent post here. Thanks to Gabe Paley for sending this along. Though, really, the city should lobby Governor Hochul (or the next governor, if someone else wins the 2026 election) to fix the system at the state level. The Mamdani team could tee up the shape of those reforms, or the state could take the initiative. These things are fixable. Our elected leaders just need to decide they are a priority.

I wrote this a year and a half ago, before DOGE put the federal workforce in the headlines. I’m sharing it now with state and local government in mind, especially New Jersey, Virginia, and New York. all of whom have new leaders-elect. New administrations have a unique opportunity to reimagine not just how they hire and manage the workforce, but the work those people actually do, and to reengineer every system its people rely on to better balance the go energy against the stop energy. Whoever does it best wins. Game on.

From last April:

******

NextCity recently published a hot take by Steffen Berr tying the ways in which the US is failing at reducing pedestrian deaths to the misaligned training that most transportation engineers in the US receive. Berr explains that a transportation engineer “is a really a civil engineer who has received a little exposure to the transportation sector.” Due to the structure of accredited degree programs, “In a best-case scenario, a civil engineer will only take three transportation classes during their bachelor’s degree. In the worst case, they’ll only take one: Introduction to Highway Engineering. To put this into perspective, the most educated professionals we entrust to design and run our roads and streets have received only half of a minor with a handful of credits on the topic.”

Berr goes on to address the reasonable objection that in many fields, people learn on the job. But what transportation engineers learn on the job, per Berr, is not things like how to choose the most appropriate intersection for the desired use, how the road system should be laid out at a network/route level, or how to fix congestion (none of which, he argues, they learn in school either.) Instead, they learn “how to navigate the impressive amounts of bureaucracy that have been built up in the industry, memorize an impressive vocabulary of technical jargon, practice with design software like AutoCAD to produce engineering plans, and how to copy the current engineering standards. There is no exposure to deep levels of theory that can help our future professionals create original solutions to fundamental problems like safety, congestion, emissions and ethics.” 

I’m less interested in Berr’s point about the wrong degree requirements than I am in his observation about what the job of transportation engineer actually is. As Stafford Beer observed, “the purpose of a system is what it does,” and by analogy, the purpose of a job is not its stated goals but what the people who do it actually do day to day.1 When talking to people who’ve never worked in government, the biggest disconnect is usually a lack of understanding of the actual jobs of public servants. A rather dramatic illustration of this comes from a Mercatus Center podcast with Lant Pritchett in which he shares an anecdote about advocating for evidence-based policy in the Indian bureaucracy. 

After they had done the RCT [randomized control trial] showing that this Balsakhi program of putting tutors in the schools really led to substantial gains and learning achievement and reading outcomes, he took it to the secretary of education of the place in which they had done the RCT. And he said, “Oh, by the way, I have the solution to your problem of low learning levels, or at least part of the solution. Look, we’ve got this powerful evidence that this works to improve leading outcomes by putting these volunteer tutors and pulling their low learning kids out.”

The response of the secretary of education was, “What do you think my job is? Why do you think that this is a solution to a problem I have? Look around my office. See these piles and piles of files that keep me busy 60 hours a week and not one of these files is about a child not learning. I’m under no pressure about that problem. If I try and transfer a teacher, I’ve got a court case on my hand. If I try and close a school, I got a court case on my hand. My job is to administer the existing education policy such that there’s policy compliance. Super kudos to you for this cute little study you’ve done. It has nothing to do with my job as secretary of education.”

Ouch. And that’s a secretary of an agency serving a country with 1.5 billion people. 

Subscribe now

I suspect a lot of public servants in the US will read that and think “My job is not quite as bad as that but it sure feels that way a lot.” The people I know maintain enough connection to the actual mission to avoid such a meltdown (though I find the secretary’s frankness refreshing.) But both these stories help explain a conundrum that many who care about effective government (or, shall we say, state capacity) struggle to explain: the contradiction between the dedication, smarts, and creativity of most public servants and the sometimes terrible outcomes they are associated with, like the recent tragic lapses in administering student loans by the US Department of Education. (Or in Berr’s world, the 40,000 traffic deaths we’re stuck with every year while countries like the Netherlands have dropped their own already low number by 46%.2) To be sure, there are often extraordinary outcomes, and we notice them far less often, to our own detriment. But while it’s impossible to give government a meaningful overall grade, if its job is to meet challenges we face (national security, climate change, an effective safety net, etc.), we are at risk of falling dangerously short. The problem isn’t that public servants are doing a bad job, it’s that they’re doing a great job — at the wrong jobs.

The (unnamed in this context) Indian Secretary of Education seems to agree: “My job is to administer the existing education policy such that there’s policy compliance.” I highly doubt that’s the job he thought he was getting, or the job he wanted to do. Berr is on the same general theme when he says that what transportation engineers learn on the job is “how to operate in the industry effectively as it has been currently set up.” Note his use of the word effectively. Effective towards what? Not towards reducing traffic deaths or congestion levels. “All the experience in the world of copying and pasting a standard invented fifty years ago is useless when the problems that the standard was invented to resolve have changed,” he says. “Understanding this sheds a lot of light as to why 40,000 people are still dying on our roads every year and why your local city insists on laying down sharrows [which are known to be ineffective and often dangerous] in their latest round of “safety improvements.” Quite frankly, it’s because we have no idea what we are doing.”

This is a useful nuance as I develop a framework for building state capacity. One of my admittedly obvious and oversimplified tenets is that systems have both “go energy” and “stop energy,” much as a car has a gas pedal and a brake. You wouldn’t drive a car without a brake, but you also wouldn’t drive a car in which the brake was pressed all the time, even when you were trying to accelerate. This is a good metaphor for how we’re dealing with the implementation of CHIPS, IRA, and the Infrastructure Bill, for example, where the clear intent is speed and scale but the public servants responsible are held back from that by the brakes of overly zealous compliance functions. I hear a version of this at every agency I visit: “Congress tells us to do something. Then the compliance offices keep us from doing that very thing.” (And side note for further discussion: This is an issue of representation, voice, and democracy.) The stop energy in our government is currently a lot bigger than it should be. We’re hitting the gas but we’re not accelerating because we’re pressing the brake at the same time. 

Lots of people in government have “stop energy” jobs. We need them, and we need them to be good at them. I don’t want to live in a country where our government doesn’t exercise “stop authority.” I try to remember not to complain when my flight is delayed because I really don’t want to die in a plane crash, and a rigidly implemented checklist is a big part of how we keep safe (the current epidemic of doors and engine cowlings blowing off notwithstanding). I also really like being pretty confident that a pill I’m taking has been tested and not tampered with. I like thinking our nuclear arsenal is protected. You know, little things like that.

Stop energy is critical. Rigid adherence to protocol is usually lifesaving. But it must exist in balance. I recently learned the Navy concept of “front of sub/back of sub.” The back of a nuclear submarine, where the nukes live, is run by the book. You don’t deviate from the checklist. You don’t innovate. You don’t question. The front of the sub, on the other hand, is responsible for navigating through dark waters. You have to improvise. You have to make judgment calls. There are manuals and checklists, for sure, but the nature of the work calls for a different approach, and the Navy recognizes that the cultures of front and back have evolved appropriately to meet distinct needs. 

There are times, of course, when you’ll need front of sub judgment in a back of sub context. If the plane I was on was about to be bombed by an enemy combatant (unlikely in my life, I hope), I would be okay with the pilot using her discretion to cut a corner or two on the takeoff checklist, because the very thing that checklist is there to protect (the lives of the people on board) would under threat from a different vector. Taking every precaution in that scenario could be reckless. That’s a bit how I feel about the NEPA reviews and other bureaucratic processes that are holding back building the infrastructure we need to move to a low-carbon economy. I wish for the public servants in charge to see the threat of inaction – those species the checklist is trying to protect are threatened by temperature rise as much or more than they are by the project in question – and make good judgment calls about getting the plane off the runway a lot quicker, so to speak. This feels like a domain where back of sub culture has more hold than it should given the circumstances. And to Berr’s point, we can’t rely on back of sub culture when the checklist and protocols it uses no longer serve the purpose.

Of course, “stop energy” roles can themselves be balanced – if only I had a dime for every discussion about the value of lawyers who get to yes and the frustrations with those who seem to do nothing but block. The analogy breaks down a bit here because the items on a pre-flight checklist are binary – they are either red or green – whereas the ad hoc checklists that lawyers assemble to ensure compliance before signing off on an action are almost always shades of gray – they can be open to lots of interpretations. Any given lawyer, or compliance officer, or ethics cop can treat their role with appropriate balance, reserving their stop authority only when the risks truly outweigh the benefits. But getting the culture of a team, department, or agency to balance stop and go correctly at a macro level is extremely difficult. It’s rare to see leadership really change that balance, or for it to stick. It’s a retail approach, hugely dependent on personalities and circumstances.

What would a wholesale approach to getting back into balance look like? One answer should be a simple matter of top-down workforce planning, of the kind our Office of Personnel Management should be empowered to do: fewer stop energy jobs relative to go energy jobs. Hire more doers than brakers, both in how the position is defined and in the characteristics of the people selected for the job. But that proposal needs several important caveats. Of course, every great employee is some mix of these energies – a “go only” employee would be exhausting and dangerous in all but the most extreme circumstances – so we’re talking about a general orientation. More importantly, having fewer brakers will only result in enormous backlogs if they have the same stopping power. But there are plenty of functions where its possible to safely move from default no to default yes, possibly with an after the fact correction mechanism.3 Instead of requiring form redesigns to go through a long White House approval process before they can be made available to the public, for instance, allow agencies to apply the appropriate level of scrutiny and sign-off for the form at hand and develop a process for catching and quickly fixing anything determined to be detrimental. This example speaks to the issue of multiple levels of safeguards. Loosening a safeguard that operates at the top level of federal government may not make much difference to overall stop energy if agencies, or in turn their subcomponents, or even teams, react by strengthening their own safeguard processes. There might be something like a Law of Conservation of Safeguards at play here. But it’s still worth considering the value of moving to default yes processes where appropriate. 

Share Eating Policy

Of course, the question of the nature of the job public servants are tasked with is about much more than just stop vs go. It’s about what kind of work we’ve decided to invest in. I go into some depth about this in Chapter 5 of Recoding America as it relates to our lack of investment in digital competencies and how ideologies about private sector superiority led to a big outsourcing push just as digital was beginning to massively transform society.

…these internal competencies in digital became necessary just as we were jettisoning internal competencies of all sorts, not developing them. Instead of digital competency, government has developed extensive processes and procedures for procurement of digital work, and the ins and outs of procurements sometimes seem more complex and technical than the latest programming languages.

This points to another way to understand the disconnect between high employee performance and the outcomes our government produces (or fails to), especially relative to the investment made.4 Take procurement. I know a lot of people in procurement who are really good at their jobs. Some of them are considered really good because they’re great at the “back of sub” tasks of making sure every box is checked, and a manager might feel compelled to give them a high performance rating because of their thoroughness and dedication, even if the people who need the thing being acquired are frustrated by the slowness and rigidity of the process, and even if the thing that is ultimately acquired has checked all the boxes but doesn’t actually work. (For an example of this, see Chapter 4 of Recoding America.) But many of these procurement professionals operate according to “front of sub” principles, and are enormously creative and mission-driven. The other public servants who rely on them to procure things value them enormously. They may or may not receive high ratings, if the manager is judging them based on a “back of sub” approach. But procurement processes simply should not be as complex and burdensome as they have become. Both of these kinds of procurement professionals are doing a job that simply shouldn’t exist in its current form.

Especially with the looming threat of the return of Schedule F under a possible Trump administration, there’s a lot of talk of public sector employee performance and protections. I agree strongly with Donald Kettl, who has said about the left’s silence on civil service reforms in the face of Schedule F: “You can’t fight something with nothing.” I hope to be part of proposing a something there, something that improves government’s ability to fill many open positions and to effectively and ethically manage the workforce. But we could succeed entirely at that and still fail to meet the challenges in front of us if the jobs we fill are the wrong jobs.

Another of my admittedly obvious and oversimplified principles of how to build state capacity is that there are really only three things you can do:

• You can have more of the right people

• You can focus them on the right things

• You can burden them less.

There is obviously quite a lot to say about each of those things, and they are all deeply intertwined. A big reason we don’t have more of the right people is that we overburden both the people responsible for hiring and the applicants, focusing both on the wrong things. We overburden public servants generally because we have designed too many of their jobs to stop bad things instead of to enable the things we desperately need. We are too often asking if public servants are doing a good job instead of understanding and questioning the nature of the jobs they’ve been hired to do. 

We need a much more robust understanding of how to fix the problem of hiring the right people to do the wrong jobs. We need wholesale strategies for tuning the dial between front of sub and back of sub, between stop and go, between brake and gas, and refocusing the job of public servants on the work that’s most directly meaningful towards the outcomes we want. We need staffers in agencies who act as if the climate crisis is the enemy plane that’s about to bomb us. We need transportation engineers whose actual job – as practiced on a daily basis, at scale – is to reduce congestion and pollution and improve and save lives. We need Secretaries of Education who have time in their day to look at the study on improving learning achievement, and maybe even take action on it. We need all of this now.

Imagine a world in which this — not just enforcing rules, not even just helping agencies fill open jobs, but ensuring that federal government fills the right jobs — was the mandate of an empowered and deeply collaborative Office of Personnel Management. They couldn’t do it alone, of course — it’s agencies that define the jobs they think they need and Congress that throws down law after law they must comply with, feeding the need for compliance. The White House Office of Management and Budget adds its own reporting and compliance burdens. Each would need to buy in on an agenda of building state capacity and do their part. But this is what workforce planning should really be, and in 2025, we will need it more than ever.

Share

Share

If you’ve followed my work, you know that I spent about ten years working on what was first government technology, and then became about government service delivery, and ultimately became just about reforming government. Each shift took me upstream of the problem I’d been previously focused on, and taken together, they revealed two fundamental challenges. The first was that it is incredibly difficult to shift how government builds and buys technology without shifting the other aspects of its operating model. The second was that the conditions for transformational change were fundamentally not there. A lot of good and valuable change has happened over the past decade and a half, but reform efforts had to swim upstream, leaving most of government still wildly over-proceduralized, expensive, and ultimately struggling to achieve the goals it sets – and getting in the way of worthy aspirations of non-government actors in the private and non-profit sectors. The time was not right for those big, systemic changes to take root.

Today, I believe it is. Not only are we facing shocks from all sides — budget and workforce cuts at all levels of government (most visible in the federal government but also for states, counties and cities), the impact of AI adoption in both the public and private sectors, DOGE moving the Overton window for the speed and scale of change — we are also finally reckoning with the fact that the government we have today is no longer fit to the work we need it to do, and too often it’s an active impediment to that work getting done. Whether you like it or not, disruption is here. The job now is to shape that disruption in the public interest.

That’s why we’re launching the Recoding America Fund. It’s a pooled philanthropic effort that will raise and deploy at least $120 million over the course of the next six years. We’ll build off the work of so many over the past decades who’ve seen this problem from a variety of angles: the Congressional modernization field, good government groups, the civic tech community, think tanks who see the futility of their policy recommendations when they land in a system that can’t faithfully execute them, private sector leaders and contractors who have had to grapple with unnecessary government bloat, and others. But while this work will leverage the powerful insights so many have gained from grappling with a system in disrepair, it is also discontinuous with the past in three ways. First, the frame is no longer transparency or accountability or modernization, but rather state capacity – simply, the ability of our government to achieve its policy goals. Second, we seek to build a diverse but coherent field around that frame, one in which advocates can achieve more collectively than each could alone, while maintaining their own perspective and tending to their own politics. And lastly, we are collectively committed to leveraging this moment of disruption to achieve what’s needed for the future of our country, not what’s comfortable or convenient.

Subscribe now

To quote my mentor Mike Bracken, our task isn’t complicated, it’s just hard. We need to update the operating model of government. We inherited a system built in the industrial era to win a world war and recover from an economic depression, and then lawyered up starting in the sixties to correct for abuses of power. We still have to be ready for conflict, build an economy that works for everyone, and protect our most vulnerable, but the world around us has changed. We dealt with the advent of the information age by buying some computers, but left many processes paper-based. In the Internet era, we put up some websites, but retained the basic assumptions of industrial era management. Now we are hurtling into the AI era with an aging operating model onto which we’ve tacked some trappings of modernity. Today, we need to be clear-eyed about the failures of the past and leapfrog our operating model into one fit not for the past decade, but for the next one.

We have a view on how to do this. If you want a government that can achieve its policy goals, it must be capable of four key things:

• It must be able to attract, select, develop, retain, and manage the right people, which means we need civil service systems at all levels fit for that purpose. Our civil service systems run the gamut from merely outdated to fundamentally unfit for this critical work.

• It must ask those people to do the right work, which means reducing the procedural bloat that keeps most civil servants from focusing on the substantive outcomes the public needs and expects.

• It must have purpose-fit systems, which means changing how we build and buy technology so it can adapt to changing needs and drive effective operations.

• And it must employ test-and-learn frameworks, which means being able to act on what we learn as we pursue policy goals rather than stay locked into decades-long plans even as we can see they are not working. This also means constructing feedback loops between the makers of law and policy and those who implement them or experience their consequences, intended and unintended.

Just as Maslow’s hierarchy of needs tell us that you can’t have self-fullfilment if you’re not fed, housed, and clothed, we must address the basic needs on which all policy outcomes depend. The Recoding America Fund will resource and coordinate efforts to update each of these four elements of government’s operating model to be fit for purpose.

Fit-for-purpose means fit for not just today, but for tomorrow, and highly adaptable in a fast-changing world. As my colleagues and I have been developing this fund, I’ve had to check myself from time to time as I noticed my own desire to take another swing at the problems that have vexed me over the course of my career. Those problems remain, but the field they play out on is changing, and it will do us no good to fight last decade’s battles. The fight to build and buy purpose-fit digital infrastructure, for instance, must recognize that AI changes not only the tools government can use but also the world it must respond to, raising the expectations of the public, increasing the sophistication of bad actors, and requiring even faster adaptation to changing needs. (For more on this, I recommend Dan Munz’s The end of civic tech’s interface era and The Agentic State, by Luukas Ilves and others.) AI will also change the nature of the work that government must do, and therefore the needs that civil service systems must meet. We must be vigilant about skating to where the puck will be, so to speak. Success can’t be dragging government boldly into the 2010s.

We embrace the political realignment that has made state capacity not only a prime time issue, but a common cause for what might in the past have seemed strange bedfellows. We are committed to working not only across the aisle but with leaders who have a wide diversity of ideologies that recognize the importance of this work and no longer tolerate the status quo.

Our excellent team reflects that diversity. We have veterans of Republican Congressional offices and the House Oversight Committee on our team, including Lauren Lombardo and Laurent Crenshaw, as well as former Administration and Hill officials from the Democratic side, including Victoria Houed, who leads the team’s operations. And we have plenty of help from civil servants as well as thought leaders on issues of state capacity, including Santi Ruiz, of Statecraft fame, who serves on our board. Statecraft has fast become a mandatory listen for anyone interested in the issues the Fund will seek to tackle. In addition to me and Santi, Charlie Anderson, the EVP for Infrastructure at Arnold Ventures, and Galen Hines-Pierce, an independent researcher and philanthropist focused on democratic modernization and innovation, round out our board. Galen and Charlie have been thought partners for more than a year in developing this fund and the work, and Arnold Ventures has seeded a lot of the early progress upon which the Recoding America Fund will be building.

We are fortunate to have Robert Gordon leading our state-level work, which he has been developing and already executing since the beginning of this year. Robert has held senior roles in New York City, for the State of Michigan (as the Director of the Department of Health and Human Services), and at the federal Department of Health and Human Services, the Office of Management and Budget and the White House Domestic Policy Council. He has also spent time in the private sector at the College Board and in academia. States of every political shade need change; if they achieve it, they can not only help their residents, but also provide models that scale nationwide. We’re lucky Robert will continue leading and growing our work to support them.

Share

And as we launch, I’m thrilled to welcome our newest team member, Anne Healy, who is joining us as CEO after a rigorous and expansive search. Anne comes to us most recently from USAID where she held a groundbreaking non-political position as director of the Office of the Chief Economist. (For more about this office, I recommend Santi’s excellent interview with Anne’s former boss, Dean Karlan, on Statecraft. As he talks about moving from evaluations based on accountability to those based on impact, you’ll recognize some of the principles that drive our work.) Anne previously led AID’s Development Innovation Ventures and has held key roles at J-PAL, Evidence Action, and the State Department. Anne is passionate about the need to reform the administrative state, and we’re delighted to have her leadership on this critical effort.

We’re also delighted to have the support of some of today’s most forward-thinking philanthropists, who have helped us achieve one-third of our fundraising goal thus far. (We are still fundraising, and welcome support!) Arnold Ventures, founded by Laura and John Arnold and led by President and CEO Kelli Rhee, is anchoring the fund, and we are inspired by their novel thinking and dogged persistence on so many critical issues. Those characteristics also describe Seemay Chou and Jed McCaleb, who are major contributors through the Astera Institute. They fund transformative science, but support Recoding America because they believe that whatever your policy goals, the road to scale runs through government. The Packard Foundation, Alta Futures, RMR Foundation, Minerva Fund, Steve Newman, and John Wolthuis are also supporting, with several others in the process of finalizing their commitments, including the Hewlett Foundation and other major donors. We are grateful to all our donors for recognizing this critical unmet need.

This fund exists because a small set of people who have been doing the work saw the need to connect and supercharge their efforts with others. When Kumar Garg, Parth Ahya, and Anjali Fernandez at Renaissance Philanthropy originally ganged up on me at a conference, convinced that a fund was needed, I was staunchly opposed. But Daniel Correa at the Federation of American Scientists, Ben Bain at the Niskanen Center, and Charlie Anderson at Arnold Ventures brought me around. It turns out that Galen Hines-Pierce, Marci Harris at POPVOX Foundation, Zach Graves at the Foundation for American Innovation, and Travis Moore at Tech Congress were thinking along the same lines, and we joined forces. This group, as well as Jenny Mattingley at the Partnership for Public Service and Alec Stapp and Arnab Datta at the Institute for Progress, and a few others have served as the core of our braintrust, informing the strategy and sketching the outlines of both a new operating model for government and the path to get there. As we grow, we’ll be tapping a wider range of partners and grantees.

And we’ve assembled an impressive board of advisors to bring a diversity of perspectives to our strategy. Together they bring experience from the state and federal levels, from the executive and legislative branches, from a birds eye view and the nitty gritty details of implementation, and of course, from the left, right, and center.

• Ben Buchanan, Dmitri Alperovitch Assistant Professor at Johns Hopkins School of Advanced International Studies and former White House Special Advisor on AI

• Dan Ho, Professor, Stanford University; Director, Regulation, Evaluation, and Governance Lab (RegLab); and Senior Fellow, Stanford Institute for Human-Centered Artificial Intelligence (HAI)

• Daniel Stid, Senior Fellow at the American Enterprise Institute

• Dean Ball, Senior Fellow at the Foundation for American Innovation, and former Senior Policy Advisor for AI and Emerging Technology

• G.T. Bynum, Vice President of Community & Government Affairs at Saint Francis Health System and former mayor of Tulsa, OK

• Jeanne Lambrew, Director of Health Care Reform and Senior Fellow at the Century Foundation and former commissioner of the Maine Department of Health and Human Services

• Judge Glock, Director of Research and a Senior Fellow at the Manhattan Institute

• Marc Dunkelman, Senior Fellow, Searchlight Institute; Fellow at Brown University’s Watson School for International and Public Affairs; and author of Why Nothing Works

• Marina Nitze, Director, Child Welfare Playbook and co-author of Hack Your Bureaucracy and Crisis Engineering

• Marshall Kosloff, Host of The Realignment and Arsenal of Democracy Podcasts, Fellow at the Clements Center for National Security, Foundation for American Innovation, and the Hudson Institute

• Nicholas Bagley, Professor of Law at University of Michigan and Senior Fellow at the Niskanen Center

• Reeve Bull, Director of Virginia Office of Regulatory Management

• Solitaire Carroll, former Acting Director of Veterans Experience Services Portfolio at the U.S. Department of Veterans Affairs, now writing on digital government for Niskanen Center

• Stefanie Sanford, Chair, Institute for Citizens & Scholars and President of Civic Ventures at Alithi Consulting

• Zachary Liscow, Professor of Law at Yale Law School, former Chief Economist at the Office of Management and Budget at the White House

We know this won’t be easy. Updating government’s operating model means confronting decades of accumulated procedures, outdated systems, and entrenched interests. But the convergence of forces we’re seeing today—technological disruption, fiscal pressures, and political realignment—has created an opportunity for change that we must take. Whether your priority is national security, economic and technological dynamism, protecting the vulnerable, or simply getting basic services to work, none of it is possible without a government capable of executing on its goals. The time for incremental fixes has passed. Let’s build the government our future demands.

Subscribe now

Want to get involved or just follow along? Sign up for updates at the bottom of this page.

You’ll need a bit of the back story. As you may recall, in December 2023, the Free Application for Federal Student Aid (FAFSA) launched its new system, three months late and with what the Department of Education would later call, in classic bureaucratic understatement, "significant and widely-reported challenges." What they meant was that millions of students could not access financial aid.

Subscribe now

So far we’re in quite familiar territory. The FAFSA had run the play many agencies still follow: vendors working in isolation, little internal technical expertise, a big-bang deployment, and a detailed set of specifications written with little understanding of the actual need. It was a perfect example of what I’ve called the project model, governed by rigid, waterfall project management, which amounts to a pledge by all parties not to learn anything while doing the actual work, to quote Clay Shirky. The predictable result (predicted, in fact, by the department’s own staff) was that the FAFSA Processing System had "availability issues, recurring errors, and long wait times." Forty percent of calls to the help center were dropped before anyone even tried to help. Students couldn't get aid. Colleges couldn't plan. Parents saw their dreams for their kids slipping away.

But in 2024, Ed began to right the ship. This year’s FAFSA tells a dramatically different story. Through July 2025, more than 14 million students completed the form—11.4% growth compared to the previous year. Student satisfaction hit 93%, with 88% saying the form took a reasonable time to complete. The National College Attainment Network called it a "tremendous achievement."

What happened?

Enter the Product Model

I've written before about why government needs to move from project-based to product-based operating models when it comes to technology, and my janky graphs illustrating the difference apparently got passed around quite a bit. Then Ann Lewis and I laid out the case for the product operating model in detail in our Niskanen Center paper. The basic idea is simple: instead of treating technology as a one-and-done project completed by deathmarching through a colossal set of predetermined features, you create persistent teams that can act on what they learn as they go and measure success through delivery of meaningful outcomes. In order to do that, we said in the paper, you have to align several elements: funding mechanisms; team structure and roles; roadmapping and prioritization practices; decision-making authority; and success metrics. But we missed one. You also have to align your oversight bodies.

Share

What happened is that the department brought in a team from the College Board with a long history of working in the product model, and they have transformed the FAFSA and the team responsible for it. Starting last summer, this small band of product leaders, with help from a few detailees from the US Digital Service, took the reins and not only wrestled the form back into functionality but also steered the entire enterprise in a new direction. (Steering makes it sound a lot easier than it was, but you get my drift.)

The new team started hiring actual engineers and product managers. They started procuring capacity from vendors instead of specific deliverables, enabling quicker responsiveness to bugs and user needs. 55% of contractors’ entire FAFSA team used to be dedicated to manual testing, so they automated testing, which enabled them to move from 67 testers to 32, while at the same time dramatically reducing "regressions" where new features break existing functionality. They started doing smaller software releases far more frequently, reducing the risk of introducing bugs into existing functionality, and making it easier to measure outcomes and fix issues quickly.

These changes, especially getting far more digital expertise in the FAFSA staff who could make actual decisions about priorities, allowed them to quit the death march through the predetermined feature set and respond to what users actually needed. The original contract required custom-built analytics software, but the engineers and product managers now on the team knew that commercial off-the-shelf offerings would be better and cheaper, so they just bought what they needed, freeing up budget and development capacity for actual user-facing improvements.

The contract also called for a renewal capability feature – a way for users to import their prior year answers as a starting point for their form. Clearly the FAFSA should have this feature. But the team could see that about 1 million students weren’t making it through the part of the application where they have to get their parents on the platform to input their employment and income information. That was a clear and urgent need. The team decided to fix the invite experience immediately and let the contractually specified renewal feature wait. It was the right decision, enabled by the elements of our model as described above, especially a new definition of success in which helping the students mattered more than sticking to the plan. But it was a decision a legacy IT project team could not have made. The standard combination of vendors obligated to fulfill the terms of a contract and project managers whose job it is to make sure they do just that would have locked the Department of Education into the wrong priorities, and worse outcomes for kids applying for aid.

This is what good product management looks like: user data drives decisions, not fidelity to a plan written before anyone understood the real problems.

GAO doesn’t get the memo

Fixed system, happy users, costs contained—what's not to love? Well, as evidenced by a report out earlier this week, the Government Accountability Office found a lot not to love. Because the new, improved FAFSA team hasn’t been following the same rigid, compliance-heavy, project-based approaches that had created the original disaster, GAO penned a 60-page slap on the wrist.

The report reads like a greatest hits of outdated IT management: demands for better "contract monitoring processes," documentation of "contractor performance reports," and ensuring "assessments of contractor performance are documented in accordance with departmental guidance." It's like scolding someone for putting out a fire with water instead of the gasoline that started it.

Read it if you like, but the GAO report is not today’s “obscure government documents you may find entertaining.” The thrilling read is the FAFSA team’s rebuttal. Aaron Lemon-Strauss, the Executive Director of the FAFSA Program, penned a nine-page letter that is essentially a point-by-point demolition of GAO's project-thinking. It's polite, thorough, and devastating.

The letter makes clear that GAO fundamentally misunderstood how the FAFSA system actually works. It wasn't just one contract with one vendor, but a complex ecosystem of five different components across multiple contracts. But more importantly, the letter defends the product operating model that actually saved the FAFSA.

GAO complains about the team not following rigid contract specifications, but the FAFSA team explains why that approach fails with complex software systems: "The lesson to draw from this is not that someone had made a foreseeable mistake for which they should be held accountable, but that with complex software systems it is impossible to foresee how everything will work together until you are actively building the features."

They cite our Niskanen Center report and make the case for continuous, user-centered delivery over milestone compliance. Their suggested revisions to GAO's recommendations are a clinic in how to think about government technology differently. Instead of "improve contract monitoring processes," they suggest "create time-bound Objectives and Key Results for improvements." Instead of "validate contractor performance reports," they recommend "establish a process for demonstrations of completed work after every development sprint."

They're teaching GAO how modern software development actually works.

Aligning Oversight

The FAFSA team learned the hard way that it’s not enough to adopt better practices internally. You also have to educate the entire ecosystem of people who fund, mandate, and oversee your work. Part of the job of shifting from an outdated model to one more fit for purpose is helping Congress, oversight bodies, and other stakeholders understand what they should be holding you accountable to.

Share

But leaders at GAO and other oversight bodies need to take responsibility for educating themselves about how modern technology development works and updating their playbook. They can’t keep applying 20th-century accountability frameworks to 21st-century challenges. Instead of asking whether a team delivered specific features on time and under budget, they need to ask whether they're improving user outcomes and learning from data. Everyone at GAO should read Lemon-Strauss’s memo ten times, and ask 100 questions until they understand everything he’s saying.

A final point: The Department of Education leadership deserves kudos for backing Lemon-Strauss and publishing his frank rebuttal. Far too often in government, a bad GAO or Inspector General report is the death knell of change. Promising new tools get shelved, new practices reverted, and “lessons” are “learned.” People often ask me why there aren’t more product teams in government and the answer is that many teams try to make the shift, but a good number of them end up right back where they started because someone, somewhere calls foul, and leadership buckles. Sometimes it's not an actual critical report that kills progress, just the threat of one. The final ingredient of the model, then, is the backbone to defend these practices against oversight that fundamentally misunderstands what it's overseeing.

Thank you, FAFSA team, for showing the way forward, for standing up for yourselves, and, of course, for fixing federal student aid. And thanks for the spicy memo. This one goes in the Obscure Government Documents Hall of Fame.

Subscribe now

Subscribe now

If I were writing this book, I would write about how Matt Mahan, the mayor of San Jose, responded to a constituent’s tweet about a permit. Donovan Adesoro works for BuildCasa, a startup trying to build more housing. The San Jose permitting office rejected documents his team filed because the margins are supposed to be 1 inch, but on one of the pages, it appears to drift to 1⅛ inch. (At least that’s what the picture makes it look like, maybe they’re 1 1⁄8 inch everywhere.) He posted:

Adesoro didn’t tag Mahan on his post, but the mayor replied anyway.

Later the same day, Mahan follows up;

There is so much to say about both the absurdity of the county’s persnicketyness and the deftness of Mahan’s communication here. He doesn’t just acknowledge and explain. He enrolls Adesoro in a cause, and invites anyone listening to join as well. But the real triumph here is that he engaged in this issue in the first place. So many politicians would consider this kind of administrative detail unworthy of their attention. If they’re on Twitter, it’s to put out messages about their values, their policy priorities, their visions. They’ll talk about aspirational goals. Mahan is listening, and understanding what’s blocking us from achieving the goals we’ve already set. We need more of that.

Share

More housing is a goal California has clearly set. According to the 21st Century Alliance, the California Legislature passed more than 100 bills to stimulate housing production between 2016 and 2022. Many elected officials deserve praise for these legislative accomplishments. But, as the Biden administration learned, legislative accomplishments do not necessarily ensure real-world change. Despite these 100 bills, housing permit numbers in the state have barely budged, and prices in the state remain sky-high due to low supply. Housing costs remain among Californians’ biggest concerns. Obviously, something else is needed. Perhaps there’s a learning cycle that informs and improves the effectiveness of the next 100 bills, but clearly, it’s not just that. No bill is going to change the overly rigid culture of the county and state permitting offices. For that, you need good old-fashioned leadership.

Mahan is a mayor, not a state legislator, and not the governor. (Though I do wish he would run for governor.) This kind of on-the-ground practicality is seen as somewhat characteristic of mayors, though his willingness to dive into the weeds is remarkable even for a mayor. But electeds of all stripes should take a page from Mahan’s book. And other legislators already have. I’ve praised Marie Gluesenkamp Perez’s crusade against the regulations (or interpretation of regs) that kept daycare centers from serving fresh fruit, for example. That took a similar combination of administrative sleuthing (yes, likely conducted by staff in both cases), really strong communication, and the instinct to pull a thread most leaders wouldn’t even notice.

What Mahan and MGP are demonstrating is a conception of the job voters have hired them to do that’s different from their peers. If politics is at all like gardening, most politicians think of themselves as planting seeds. If a bill they author passes, that seed will grow, and the shade, or fruit, or flowers it provides the public is what gets them renown and reelection. What they fail to see is that those seeds aren’t growing. The fruits of their legislative labors are not materializing. As any gardener knows, planting is not the hard part. Gardening is about creating the right conditions: tending and fertilizing the soil, watering, weeding to give seedlings the space they need to grow. Gardening is cultivation.

Effective leadership in today’s sclerotic government also requires cultivation. The word stems from the Latin word "colere," which means "to cultivate, to till, to inhabit, to frequent, or to tend." This Latin root is also the origin of "culture." Mahan and MGP are willing to grapple with the culture of these agencies, and ways that culture is “eating” their well-intentioned policies. They recognize that the outcomes their electorate need and expect won’t come from just planting the seeds of policy ideas if the soil those policies planted in is dry, barren, and choked with the weeds of every other past priority, requirement, and mandate of their predecessors. Instead of continuing to plant seeds and blaming someone else when they don’t grow (or worse, not even noticing they haven’t grown), the leaders we need cultivate a purpose-fit culture.

The heart of that culture work needs to be taming risk aversion and fetishization of procedure. In that culture, it isn’t acceptable to reject an application for much-needed housing over ⅛ of an inch. But the culture isn’t going to change itself. Change requires leadership. Grandstanding about the big issues of the day may grab headlines, but debugging our government an ⅛ of an inch at a time is the kind of leadership we need.

Subscribe now

As the pandemic put millions of people out of work, states developed a backlog of unemployment insurance claims. Many governors blamed COBOL, a programming language developed in 1959 and still in widespread use. But COBOL was not the real culprit. In California, where I chaired a task force to clear the state’s backlog, parts of the system using COBOL chugged along with remarkable resilience. And although about half of the states had modernized their mainframes prior to the pandemic, those states fared no better on average. If retiring legacy code doesn’t fix the problem, something else is holding our systems back from resilience and scalability.

Subscribe now

New Jersey’s Labor Commissioner Robert Asaro-Angelo put the blame where it belongs. Called to testify to the legislature about the backlog, he brought along several boxes labeled “7,119 pages of active UI regulations” and placed them prominently in view of the members. “We are putting duct tape and Band-Aids on outdated policy,” he told the legislators. The technological issues his agency was grappling with were just a symptom. The underlying cause was the complexity and sheer volume of rules and regulations computer systems were supposed to operationalize.

Asaro-Angelo’s team ultimately worked through the backlog, and analysts later called their response among the best in the nation. But the problems that make state systems fragile and inadequate in times of steep unemployment have not been solved. Until state legislators (and the U.S. Department of Labor) recognize their role in that dysfunction and start to address the statutory and regulatory complexity, they won’t be.

Regulatory cruft is not unique to unemployment insurance. In many domains, new rules are layered upon old ones; seldom is there a corresponding effort to reconcile or eliminate outdated, redundant, or conflicting provisions. This “policy clutter” creates a dense and often contradictory web that makes effective implementation difficult. With the right competencies, the executive branch can shield the public from some of that complexity, but without effective partners in the legislature, it can accomplish only so much. We need lawmakers eager to channel their inner Marie Kondo to do the life-changing work of tidying up.

Making matters worse, a different kind of policy clutter hampers the competencies and capacities agencies need to address the problem. Civil-service rules in many states constrain and prolong the hiring process, effectively turning away the necessary workforce. Procurement is no better. When our task force started work in July 2020, the California Employment Development Department (EDD) was about to request proposals for a “business system modernization,” a project the department had already been working on for 11 years. To clarify, 11 years was not how long it took for a vendor to develop the new system; it took 11 years for the state to be ready to solicit bids. It would take a few more to collect and review those bids before awarding a contract. Only then would the work start, at which point the requirements collected over a decade earlier were sorely out of date. The glacial pace is due to an inscrutable web of procurement and contracting rules. Agencies subject to such regulations are largely powerless to change them on their own.

Share

Such rules almost always start with good intentions. But their impact is often at odds with those values. In 2017, for instance, San Francisco imposed a ban on doing business with states that failed to support LGBTQ and abortion rights. The city even prohibited certain of its employees from traveling to those states on city business, even to woo companies to relocate to San Francisco.

The policy became a burden, partly because adherence was so impractical, with staff repeatedly applying for waivers to circumvent it, and partly because every acquisition, from computers to toilet paper, required paperwork proving compliance with the ban. In one case, an LGBTQ-owned venture that had been doing business with the city for years was cut from its supplier rolls when it was bought by a company in one of the 30 banned states. Prices went up, as suppliers who still qualified realized how much less competition they faced. The law was cited as one of the reasons a public toilet came with a $1.7 million price tag. When it was repealed seven years later, not only had the ban failed to move social policy in any of the banned states, its effect on city finances was disastrous and it harmed some of the very stakeholders it was meant to help. Policies like this also hurt the reputation of blue-state governance.

Conventional wisdom says the job of a legislator is to legislate. Most of the laws created for executive-branch agencies are either mandates or constraints (something they must do or can no longer do). But agencies are now drowning in these rules. As one leader put it, “It’s like I’m supposed to run a marathon, but I have to ask ‘Mother, may I?’ every step of the way.”

The best state lawmakers recognize that adding mandates and constraints is just one tool at their disposal. Because legislation doesn’t need to add to the clutter; in fact, it can be written to clean up the mess. “Clear the clutter” might not sound like a winning agenda with your constituents, but they’ll feel the effects every time they do business with the state.

As recognition of policy clutter has grown, so have the tools used to address it. Even a large and expert legislative staff can fail to grasp the intricate web of laws, policies, and regulations governing a single issue when it spans dozens or hundreds of documents and thousands of pages. But large language models (LLMs) perform impressively at these tasks.

One such model is the Statutory Research Assistant (STARA), developed by the Stanford RegLab. STARA is an automated system capable of performing accurate “statutory surveys”—compilations of all legal provisions relevant to a particular issue—along with detailed annotations and reasoning. A recent research paper by Daniel Ho and others details some of the ways his team, in collaboration with public sector leaders, have used the tool.

Subscribe now

One example examined the federal criminal code. As Ho and his co-authors write: “Title 18 of the U.S. Code is notoriously dense and complex, consisting of 1,510 sections that span tens of thousands of pages of text. While Title 18 is the main federal criminal code, there are federal offenses that are not included in Title 18.” Justice Department officials have been trying to get Congress to recognize the negative consequences of this complexity since 1982, when they first tried to count the number of federal crimes. After two years of combing through statutes, they could arrive only at an educated guess (3,000 crimes). In 1998, the American Bar Association tried its hand, concluding that the number was likely higher than what Justice had estimated but failing to return an authoritative figure. In 2019, researchers at the Heritage Foundation tried again, under the banner of “Count the Code,” by searching for phrases—“punished by a fine” or “shall be fined or imprisoned”—commonly associated with criminal offenses. This time, they confidently put forward a number: 1,510. But STARA, leveraging more sophisticated technology, recently identified 2,305 criminal provisions (53% more) with 98% precision (yes, they checked)—nearly 800 more than previously documented.

Although a necessary first step, gaining an accurate picture of the problem does nothing to reduce the complexity of the criminal code. Congress still needs to act, and there are very good reasons it should do so. When prosecutors cannot easily identify all the relevant criminal statutes, it’s difficult to ensure consistent and fair application of the law across different cases and jurisdictions. Without knowing the full scope of criminal provisions, the Justice Department may struggle to prioritize enforcement areas or to allocate resources effectively. It’s nearly impossible to train prosecutors and investigators in laws they cannot systematically identify or access. And complexity is expensive. The system would be more effective and more equitable and cost less if our elected leaders could produce a moderately simpler set of rules, one that rationalizes choices made over time into something more straightforward.

Congress may not be ready to act, but states are in a position to tackle their own criminal codes—and could start by coming up with a credible count, as the Department of Justice did. Representative Marie Gluesenkamp Perez from Washington’s 3rd district caused a stir when she highlighted how confusing and restrictive licensing regulations resulted in childcare facilities being told they couldn’t serve fresh fruit to the kids in their care. Policy clutter has become a burden across so much of modern life; here are a few areas to consider.

• Housing: Developers often face overlapping or conflicting requirements, from state-level zoning statutes to related environmental regulations. These can delay projects and drive up costs. The political pressure to act on these conditions is mounting as the conversation around abundance enters the mainstream, but action requires actually understanding this tangle of rules. Mapping state-level zoning statutes and related environmental regulations could highlight redundant or overly restrictive provisions that could be simplified or removed to accelerate housing development.

• Education: State education codes are notoriously detailed and expansive, encompassing curriculum requirements, teacher certifications, standardized-testing rules, and funding formulas. Over time, these layers of policy create confusion and administrative inefficiency, and restrict innovation in education. Identifying outdated mandates, conflicting curriculum standards, and unnecessary administrative obligations could simplify educators’ and administrators’ workloads.

• Healthcare: Hospitals, clinics, nursing homes, and assisted-living facilities must adhere to an intricate web of state health codes, safety regulations, staffing requirements, and licensing standards, which often overlap with federal rules. The complexity can result in inconsistent enforcement and high compliance costs. Removing conflicting or duplicative regulations, including harmonizing reporting requirements across levels of government, could lower healthcare costs and improve service-delivery efficiency.

• Professional and occupational licensing: In occupations ranging from cosmetology and interior design to nursing and teaching, the licensing process is typically regulated at the state level and often entails outdated or conflicting criteria. A statutory review could reveal outdated or conflicting licensure criteria, providing legislators opportunities for streamlining and consolidation.

• Insurance and financial regulation: Some states have notoriously dense insurance codes, creating a convoluted compliance landscape for companies and confusion among consumers. Removing unnecessary complexity or conflicts could improve competition and better protect consumers.

This list of policy-clutter opportunities admittedly glosses over the reality of politics and change. A certain amount of cleanup will be relatively uncontroversial, slowed only by a natural human tendency to prefer the comfort of the status quo. Other proposed changes will face self-interested opposition and will require courage, and often compromise. Navigating those trade-offs requires leadership. Here are a few tips to keep in mind.

• Explore the landscape of opportunity. As these tools mature, statutory surveys will be increasingly easy and accessible. Try running these models in a number of areas before deciding where to dive in.

• Start with test cases. As you identify policies that may benefit from simplification, start with something less publicly visible, so that you can learn outside of the spotlight. Many hot-topic areas (such as permitting) are ripe for simplification; getting some trial-and-error practice under your belt first will help you get these high-stakes cases right.

• Model comfort with LLMs. STARA isn’t yet available on the web, but teams like the RegLab are actively engaging partners and models like ChatGPT’s Deep Research are commercially available and well worth the cost. Encourage your staff to become familiar with their capabilities and try them out for yourself. Upload sections of the federal or state code; tell the model what you’re trying to achieve; ask for recommendations for statutory and regulatory change. Support pilot programs within legislative research departments that explore potential AI applications.

• Engage agencies. Encourage state agencies to recommend areas where streamlining regulations could make systems more robust and scalable. Encourage them to use LLMs toward this effort, but don’t require it. Use their suggestions as a jumping-off point for further LLM-powered inquiry to understand the full scope of regulations impacting their work, and keep lines of communication open. Support initiatives to train public servants in these skills and to bring in individuals with the needed expertise.

• Let LLMs help you think before you legislate. It’s often easier and more appealing to craft a new bill than to troubleshoot or downsize existing statutes, but the latter can be far more effective. When proposing new policies, actively consider their potential impact on existing regulations and the capacity of state agencies to implement them effectively. Ask models: How will this interact with the current layers of policy? What can be achieved through regulatory and/or practice change, without a new bill?

• Leverage user research. If your state has a digital services team, they likely conduct unstructured, observational research that can complement the insights of LLM analysis by revealing how the rules play out in real-world situations. Ask them for help with methods of understanding the real-world impact of policies and identify pain points in service delivery.

If you want to become a leader in applying Marie Kondo decluttering to your state, the first principle is to not add to the problem. I’ve lost count of how many times legislators have admitted to me their regret over introducing a bill that sounded good at the time, but only added more complexity to an already mind-bogglingly complex system, with unwelcome consequences. Legislators need to undo these mistakes, and those of their predecessors, using powerful new tools like STARA to analyze opportunities for rationalization and simplification—and, ultimately, repeal unhelpful laws and regulations.

Rationalizing existing policies can move legislatures beyond merely reacting to crises, and toward building a more responsive, effective, and ultimately more trustworthy government for the people they serve. The thousands of pages of regulations Commissioner Asaro-Angelo made visible at his hearing are mostly invisible to the rest of us. But the clutter they represent is real, and a drag on your state. Let that image serve as a constant reminder that, for a legislator in a country about to turn 250 years old, right now, less is more.

Share

Subscribe now

Federal rulemaking rarely doubles as comedy. But on July 21, the Trump Administration published a Notice of Proposed Rulemaking that could read as satire. Based apparently on months of work by the Domestic Policy Council, which culminated in a “National Strategy To End The Use of Paper Straws,” this proposed regulation is aimed at doing what it says on the tin: “Ending Procurement and Forced Use of Paper Straws” by amending the Federal Acquisition Regulation, or the FAR, which governs how agencies buy.

Lots of people seem to dislike paper straws (or at least have genuinely mixed feelings on the topic), so perhaps this will prove to be a popular move. But this ban also comes at the same time that the federal government is working hard at realizing the President’s separate goal of simplifying the procurement process with a particular emphasis on paring back the rules to “only provisions required by statute or essential to sound procurement.” Last I checked, there are no federal statutes banning the use of paper straws, and I can’t imagine how this provision could be essential to sound procurement of anything.

So how did it come to be that the same agencies advancing a welcome and much needed effort to overhaul the FAR also decided to add their own non-statutory rules into the FAR?

Subscribe now

Because the ban on paper straws isn’t actually about the straws, it’s about the cultural shibboleth straws have become. This regulation reflects a persistent governing pathology: when activists can't achieve their goals through normal politics, they resort to social engineering via government operations. For all the (warranted) criticism being hurled at Democrats for “everything bagel liberalism,” this proves that conservatives can fall for the same fantasy that merely having the government “lead by example” will make a difference.

It doesn’t work. Hijacking the government’s ordinary operations to fight unrelated culture wars needlessly politicizes our institutions, snarls our civil servants in red tape, and usually fails to achieve even those unrelated objectives.

Procurement as an Ideological Battleground

The paper straw regulation sounds silly but it represents the latest chapter in a decades-long pattern of using procurement policy as a means to accomplish ideological goals unrelated to the acquisition of “property and services of the requisite quality and within the time needed at the lowest reasonable cost.”

Just as implementation of the last Administration’s signature legislative accomplishments was layered with exhausting requirements for environmental reviews, equity strategies, and supply chain diversity plans, each administration feels compelled to add its preferred toppings to the federal operational “bagel” over time, usually after they fail to achieve their policy goals some other way.

Share

In 1988, Ronald Reagan led Congress to require that all federal contractors with contracts greater than $100,000 establish “drug-free awareness” programs. In 2009, President Obama signed Executive Order 13513, “Federal Leadership on Reducing Text Messaging while Driving” that required contracting officers to include a clause in every contract directing vendors to encourage their employees not to text while driving. In 2020, President Trump required agencies to ask contractors and recipients of federal financial assistance “to certify that it will not use Federal funds to promote the concepts” like “any individual should feel discomfort, guilt, anguish, or any other form of psychological distress on account of his or her race or sex” as part of a broader push to eliminate diversity training. In 2023, on the heels of a broader push to drive clean energy and sustainability across the country, a cross-section of White House principals issued a directive to agencies to prioritize sustainable modes of transportation for official travel under the theory that “[a]s the Nation’s largest employer, the Federal Government has the opportunity to lead by example.”

These policy fights (and the many ones not highlighted here) rarely end up accomplishing even those unrelated goals: the War on Drugs has largely failed and some employers are walking back testing entirely; texting and driving is still a huge problem; diversity training has not been ended in corporate America; and the travel industry is slowly creeping towards sustainability despite the fact that much federal travel has been curtailed by DOGE.

This tendency is not limited, by the way, to the federal government. As Jen Pahlka & Andrew Greenway discussed in their recent report The How We Need Now: A State Capacity Agenda for 2025, in 2017, San Francisco similarly “imposed a ban on doing business with states that failed to support LGBTQ rights and abortion rights, and even prohibited certain city employees from traveling to those states on city business.” This resulted in a functional ban on doing business with 30 states on behalf of the city, caused a deluge of paperwork for businesses seeking waivers, contributed to the cost of the now-infamous Noe Valley toilet debacle, and failed to have any impact on any of the states it targeted before ultimately being repealed in 2024.

These efforts contribute to the accretion of rules and regulations as each administration adds its preferred requirements while criticizing its predecessors, creating an ever-expanding web of compliance obligations that distorts the system's fundamental purpose and results in overgrown procedures that prevent agencies from achieving any of their goals. In the case of San Francisco, a city Supervisor described this trap succinctly: "I don't think we are advocating our social policy goals one bit, in fact, we're holding them back to the extent that we make San Francisco this great blue progressive city look like a basket case that can't get anything done.”

Contractors and contracting officers become unwilling proxies for policy goals they have no special competency or interest in advancing. Compliance burdens multiply without addressing underlying problems. For example, an industry group noted that between 1996 and 2024, the number of required clauses for a commercial service acquisition in the Department of Defense grew from 3 to over 100. The procurement system's core mission of value and efficiency gets distorted, and the resulting dysfunction generates political backlash that undermines the credibility of reform efforts–and incentivizes the next President to take the opposite action, swinging the pendulum back the other way and necessitating costly re-work.

A Case Study in Beverage Technology

The recent move to ban the procurement of paper straws exemplifies this pathology.

In 2021, President Biden signed Executive Order 14057, “Catalyzing Clean Energy Industries and Jobs Through Federal Sustainability,” which (among many other sustainability-focused things) directed the federal government to prioritize “Reducing Waste and Pollution,” which was operationalized by the Council on Environmental Quality to mean in 2024 that agencies should phase out single-use plastics like straws.

Then the pendulum swung back. In February, President Trump issued Executive Order 14208, “Ending Procurement and Forced Use of Paper Straws” that instructed agencies to take “appropriate action to eliminate policies designed to disfavor plastic straws issued to further Executive Order 14057” and directed the Domestic Policy Council (DPC) to develop a National Strategy to End the Use of Paper Straws. Among all sorts of zany activist-driven ideas in the strategy (like the DOJ filing suit against businesses that choose to offer paper straws or the FDA deciding paper straws were poisonous), DPC included a suggestion that the FAR Council promulgate acquisition rules on the topic.

The recent Notice of Proposed Rulemaking implements this suggestion. The rule requires vendors to make formal representations about their straw policies when bidding on federal contracts. It establishes federal performance standards for drinking straws, mandating that they have "the strength and durability of plastic straws." Vendors must certify that they do “not have policies promoting the use of paper straws or penalizing the use of plastic straws.” Contracting officers are expected to track compliance with these requirements when they award contracts and vendors will have to do the same on their end.

Subscribe now

Taken alone, this requirement probably will result in fairly minimal process and cost – though, in a remarkable piece of regulatory honesty, the notice self-consciously admits that the government doesn’t actually know how many straws it’s responsible for and therefore couldn’t do the cost-benefit analysis normally associated with regulations. It’s also on its face reasonable: the public mostly prefers plastic straws, there are legitimate concerns in the disability activist community over their availability, and it doesn’t seem like the government buys enough straws for it to really matter either way whether they’re paper or plastic.

In some senses, one more tiny clause appended to a contract doesn’t mean much in the way of delay or dysfunction–the FAR is over 2,000 pages and this regulation will probably occupy less than half a page if it is finalized. If the original policy was mostly about “leading by example” rather than materially changing broader US climate policy then it seems that the Trump Administration’s counter-example will likely have similarly slim impact on the climate.

So, if trying to use the federal government’s buying power to influence the broader economy doesn’t work and the plastic straw rule doesn’t really add much administrative burden, why should anyone care?

The Tyranny of Tiny Decisions

Because at the same time as the FAR Council is issuing new rules about straws, the same people are undertaking the first comprehensive redesign of the government’s procurement rules in decades. This effort, still ongoing, appears to be making good progress in streamlining the FAR and has been met with some very positive initial reviews by informed observers. From the outside, many of us in the management reform community are very optimistic about the possibility that the Administration will make headway in reducing unnecessary procedural bloat by doing battle with entrenched vendors and industry special interests that benefit from byzantine rules.

But for all the good work happening to grant additional flexibility to procurement professionals and programs in the acquisition process, this straw episode is a prime example of how the FAR got to be 2,000 pages long in the first place. Successive Presidents (with the support of Congress and lobbyists) continue to saddle the procurement process with rules, mandatory representations, set-aside programs for ever more granularly-defined socio-economic purposes, etc. and it has resulted in a system so complex that it does not resemble purchasing or sourcing in any other sector of the economy. Despite a stated policy to remove ten regulations for every new one, an actual policy of “one in, one out” when it comes to pruning back operational policy cruft won’t actually help make government more efficient. The FAR “overhaul” won’t achieve that goal if it just adds back all sorts of hyper-partisan clauses of their own.

This same dynamic is playing out in the personnel space, with OPM finally removing overly burdensome requirements on federal resumes…but then adding in partisan requirements for applicants to write an essay praising the President’s Executive Orders. The administration set an ambitious 80-day goal for time to hire… but then required that all new hires are personally approved by political appointees designated by the agency head, creating new bottlenecks on speedy hiring. OPM removed the requirement for Senior Executives to write essays in their applications… but then assigned them 80 hours of training about the President’s priorities. One step forward, one step back (or, in the case of loyalty tests, one hundred years back).

Public trust in government institutions has eroded precisely because people see government as a vehicle for partisan objectives rather than competent service delivery, as something that serves other people, rather than all of us. In a 2022 Pew poll that attempted to diagnose the root causes of that decline, “majorities said the federal government unfairly benefits some people over others, doesn’t respond to the needs of ordinary Americans, and isn’t adequately careful with taxpayer money.” Rebuilding that trust requires demonstrating that the government can manage basic functions effectively, whether that's processing passport applications, delivering veterans' benefits, or buying office supplies, without turning it into an ideological battle.

Share

The straw rule is emblematic of how sensible reform gets derailed and government gets politicized: smart people identify prudent ways to streamline government, but partisan activists use their own pet policies to fill the gap left over when they run out of steam to achieve them politically. Sabotage is not the intention, but they stuff so many priorities into the gears that the machine stops working altogether.

The path forward is clear but difficult. Leaders need to resist the temptation to use the operation of government for political ends and build systems that work regardless of which party is in power. We need to rebalance the brake and the gas to empower civil servants to focus on the missions Congress entrusts to them rather than spending all day tangled in compliance exercises driven by unproductive oversight. The paper straw regulation may seem trivial, but it reveals whether the government has the discipline to choose efficiency over symbolism in its reform efforts.

In the end, you can have government efficiency or you can wage the culture war through administrative policy—but you can't have both. The choice is up to us.

Subscribe now

Soap Bubbles, Jean-Baptiste-Siméon Chardin (France, Paris, 1699-1779), Los Angeles County Museum of Art

In “Bringing Elon to a Knife Fight,” which I wrote less than 6 months ago, I suggested that while many were deeply concerned that DOGE would succeed in dismantling the federal government, we might also worry about what it would mean if he failed. It’s a lot harder to change government than outsiders think. Many reformers have preceded Elon, some of them other billionaires in reform roles, others relative insiders like the team on the National Performance Review in the nineties. It would be foolish not to acknowledge the successes of these past efforts but despite them, we still found ourselves on January 20, 2025 with a maddeningly sclerotic government. That sclerosis and lack of responsiveness (always a plus when your opponent is in power, not so much when your party has the reins) set the stage for the DOGE hatchet job.

Subscribe now

So has Elon failed? It depends on what you think he was supposed to do. Basically, he’s cut: cut staff, cut contracts, and even cut agencies. On the staff front, the firings have only just begun, which means that the bulk of them will happen months after Elon’s departure, leading me to wonder how much of this is an Elon front end with a Russell Vought back end (or in the case of the State Dept, a Marco Rubio back end.1) Clearly, large reductions in force are part of the Elon playbook and what DOGE will be remembered for, but Vought’s desire to slash the workforce was there long before Elon arrived and will be there long after he has left. In other words, cutting staff is not a uniquely Elon contribution, and arguably beneath the attention of someone with his purported genius for willing disruptive products into existence.

I asked Sahil Lavingia, the DOGE engineer fired for talking to a journalist about how the VA actually works pretty well, what he thought he had been hired to do. “We were supposed to build software,” he told me. Are you sure? How do you know? He tells me that in his job interview, to the extent there was one, Steve Davis (Elon’s trusted deputy) asked him a question I’ve heard he’s asked several others: “Do you know any successful tech companies that outsource their software?”

There was pretty clearly an agenda not just to cut contracts, but to do so by bringing some software development in house, which is actually very wise — and long overdue. I know of a few teams that have quietly gotten more staff since the start of the Trump term, and are delivering better results by firing poor-performing contractors and writing the software themselves. But those teams are in the minority. For most teams, their contracts have been canceled without much of a plan. Similarly, software (insourced or not) was supposed to replace people, but the people are gone without the software. They cut the workforce without cutting the work.

This rhymes eerily with what happened during the National Performance Review, which most people will recognize as the efforts around Reinventing Government under Al Gore in the 90s. John Kamensky was on Statecraft recently, and when asked about the staff cuts in that era, which mostly resulted not in a smaller workforce overall, but rather a “dark matter version of the federal workforce,” in Santi’s words (the same workers but now off the feds books and onto the contractors’), John responded:

We were hoping agencies would simplify HR and the procurement rules, which would let them do with fewer staff. But Congress ate dessert first and cut the number of people without simplifying the rules.

DOGE has done the same. In cutting the workforce without cutting the work, they, too, ate dessert first. They also don’t seem to have built much software, whether it's to save money, deliver better service, or automate work. Why?

Share

The answer, to a reasonable approximation, is that it’s really hard to build software in government, and when the DOGE team figured that out, instead of trying to make it easier, they decided not to bother. “They’d ask me how long it would take to build something, and I’d say, well, I could get the prototype up and running in two weeks, but to do everything that would be required to ship it, it would take at least 18 months.” Sahil was referring in part to the famously lengthy and burdensome ATO (authority to operate) process that government technologists must persist through before their software can go live to the world. That didn’t scare Sahil off -– by his own account, he wanted to keep writing code. But listening to his story you get the impression that DOGE leadership just gave up somewhere along the way.

If that’s what happened, they wouldn’t be the first. It’s hard in part because at first you think, okay, we need to change some laws, but when you dig in, a lot of the constraints come not from the laws (some do), but from how they’ve been interpreted and operationalized in agencies, in a maximalist, risk-averse culture. You’re left wondering where the change really needs to happen: in statute? in regulations? in guidance? In memos? Is this just risk averse compliance officers and someone just needs to knock some sense into them? The answer is sort of all of the above, in ways that can break your brain and suck your will to live. The work is doable, but way harder than it looks.

DOGE wouldn’t be the first to be daunted, but they had the least reason to be. DOGErs serve a Republican president with a Republican House and Senate. In the beginning, enough moderate Dems were willing to give them the benefit of the doubt that to the extent that statutory change was needed, they had a good chance of getting it. To the extent that compliance officers just needed to be convinced to interpret guidance more reasonably, DOGE had that power in ways that reformers working under Democrats, always conscious of propriety, did not. (See Jake Sullivan’s “self-deterrence” comment.2) They had, for better or worse, Big Balls, and big balls. Except they didn’t. Elon may have brilliantly disrupted the auto and space industries, but he’s leaving DC with the status quo, at least as it relates to technology and service delivery, largely intact.

DOGE was supposed to be about efficiency. Cutting jobs without cutting the work isn’t efficiency, it’s just chaos. In the private sector, it might work to assume that if there are half the people, they’ll find the most important work to do and let the procedural bullshit fall by the wayside, but in government a lot of that procedural bullshit is Congressionally mandated, or at least some version of it is. This leaves people like me in the surprising position of having wished that Vivek Ramaswamy hadn’t been banished from DOGE for being “annoying,” since Mr. Deregulation at least understood that sustainable change comes from changing rules and procedures, not just ignoring them. One gets the impression that Vivek’s preferred strategy clashed with Elon’s because it recognized that this was going to be harder than it looked, and Elon didn’t want to hear it. Yes, our representative democracy is annoying. Ask anyone who’s ever tried to get something done in government. But that's the game. DOGE had a strong hand, and could have played it.

What DOGE should get credit for is moving the Overton window for civil servants on risk aversion. Some of DOGE’s work has been marked by enormous (and sometimes heartbreaking) carelessness, but in an environment that’s too often been careful to the point of negligence. Now I hear stories of teams who would previously have crossed every t and dotted every i pushing to JFDI, so to speak, in ways that balance the need to adhere to our nation’s laws with the need to deliver for the American people. As much as I wish the impetus had been different, the bureaucracy needed a bit of a push, and I hope those who read DOGE as a signal to move a bit quicker will win out over the reactionary forces.

Obviously, if we’re evaluating the Trump administration’s overall impact, it’s way too early to tell whether whoever fills the DOGE void will meaningfully dismantle the barriers that need dismantling, or just continue with the chainsaw approach. They would have done everyone a favor, for instance, if they’d used their power to lobby Congress to change the rules around reductions in force, which require a “last in first out” approach that has resulted in firing some of the people federal government has needed most. But Elon has called it quits with the job not even close to done. So have his top advisors, though the team itself remains. So was I right when I predicted that the world’s richest man would meet his match in government reform? Largely, I think I was. The mistake I made was assuming he would actually try.

Subscribe now

Where that leaves me, at least, is with my eyes on the work still to be done. Everyone’s talking about the rebuilding that will eventually need to happen. The workforce will need to be rebuilt, indeed, especially if the layoffs that are coming in the fall cut as deep as they are planned to. But rebuilding right requires some (thoughtful) dismantling that hasn’t happened yet. You simply can’t restaff federal agencies under current hiring practices, for example. They’re a shamefully over-proceduralized, inefficient, and anti-meritocratic kabuki dance that serves no one, and DOGE has done little about it. (OPM, to be fair, recently issued some guidance with steps in the right direction, though it’s mixed in with some ethically questionable and probably pointless patriotism tests.) We’ve cut staff but left low-value processes like Paperwork Reduction Act compliance intact, so those left can’t choose the most important work. As DOGErs found out, it still takes many months to get an ATO before you can launch a website. The bad news is that the work of right-sizing those burdens is undone. The good news is that someone else could still do it. Republicans and Democrats should both be jumping at the chance.

Stay tuned for more thoughts on the road ahead.

Share