Philanthropy’s Willie Sutton Problem

Third wave philanthropy will need to work with government, but it will also have to work ON government

Nan Ransohoff recently blew the philanthropy world’s collective mind with some observations that should have been clear to anyone willing to do a little math. Looking at three capital pools — the OpenAI Foundation (26% of OpenAI, roughly $220B), Anthropic’s seven co-founders (who’ve pledged 80% of their wealth, ~$90B), and Anthropic employees’ donor-advised funds (~$60B) — she concludes this could mean $37-100B per year in new intended philanthropic spending, half from a single foundation, and a 12-18 month window in which the ecosystem to absorb that funding gets shaped.¹ She calls this the third wave of philanthropy, and clearly has high hopes for the impact it could have, but rightly worries about a host of rate-limiting factors that could mean it doesn’t get deployed as quickly or effectively as it could. By her math, deploying $50B/year at typical grant sizes would mean 50,000 new grants per year — I’m exhausted just thinking about the paperwork — and require something like hundreds or thousands of new “philanthropic startups” plus an Alphabet-worth of employees to staff them. Infrastructure like that does not appear overnight.

That summary sells this brilliant essay short, and I will return to some of her more nuanced insights in a bit, but I want to talk about a piece of Nan’s framework that’s missing, or perhaps just understated. Her ecosystem map has three players: funders, allocators (intermediaries who pool and deploy funds towards specific goals), and builders of the new organizations that will absorb some of this funding. She’s not totally writing off existing organizations, but her focus is on what she calls philanthropic start-ups: “high-ambition, talent-dense organizations designed to solve important public problems with the speed, intensity, and execution of a top technology startup.” She’s right that some of the AI wealth may be skeptical of anything that looks like it learned to operate in a previous era and hasn’t reinvented itself. The world has changed, and it is about to change even faster, and old solutions to new problems may well be a waste of money. She’s also right that allocators will have to play an important role here. For the big challenges this money should tackle, if every donor hires their own program officer to go deep on the complex, adaptive systems in question, it will mean that these dollars sit around waiting for infrastructure to be built and decisions to be made at a time when they will be desperately needed in the field. They’ll have to trust in intermediaries like Coefficient Giving, which has already earned the trust of a lot of tech money.

Where the money is

But even $650B a year in total future spending (Nan estimates between $37B-$100B in new AI funds, but uses $50B new on top of a $600B baseline throughout the essay) is not that much from the vantage point of what Nan might have named as the fourth player here: government. $650B is only about 5% of total government spending in the U.S. (Federal, state, and local together spend about $12.5T). Nan is arguing that this new wave of philanthropy will have ambitions to transform systems in a way that the previous wave has not, so if you count just the approximately $50B in new spending that she’s focused on, that’s less than half a percent of the public sector in the US.² Put another way, all of this new money put together could fund Medicaid for about three weeks.

In most likely scenarios (some will disagree with this!), the vast majority of our resources for the public good are going to continue to flow through government. That means that truly transformative change will require no longer accepting government’s current dysfunction as a known constraint. I assume Nan agrees with that; she mentions both the Institute for Progress and Coefficient Giving in the post, both of whom work on state capacity issues as part of their portfolios. But a plausible theory of change for getting government back on track is essential to the impact at scale this wave of philanthropy hopes for, and yet goes unnamed. Likewise, Dario Amodei, one of the likely giants of this third wave, recently published a very thoughtful essay in which he acknowledges the “slow and rickety policy apparatus” and the problems that creates but then proceeds to suggest what that apparatus should do without suggesting how it might get less slow or rickety. He understands that philanthropy will need to work with government, but fails to acknowledge (at least explicitly) that it will also have to work on government.

The thrust of what Nan sees the third wave of philanthropy doing differently is that it will eschew the non-profit industrial complex in favor of a “Silicon Valley for public goods.” She’s right that the organizations that will absorb this money are going to have to move at AI speed, but if the people funding them replicate Silicon Valley’s aversion to government, that’s going to limit the scale of the impact they can have. When Willie Sutton was asked why he robs banks, his answer was “because that’s where the money is.” Government is where the literal money is (money that could often go much further if we fixed what’s wrong with government), but it’s also where the impact is. And it’s where the legitimacy is. When government can’t act at the speed of the technology, the consequential decisions still get made, but by the companies building it rather than the public’s representatives. Many AI leaders would prefer not to be the ones holding that responsibility. Many members of the public would agree.

Why it's felt unfixable

Leaders of the previous wave have largely focused on extra-governmental approaches to solving problems not because they don’t care about the public sector but because they worry it is unfixable. When they’ve tried to work with government in their businesses, it’s been frustrating, even maddening, as time scales and ways of thinking about problems just didn’t match up, and they brought that perspective into their philanthropic work. I am no apologist for government — if you’ve read my work you know that I believe the window for modest incremental improvements has passed and that government needs a major reboot, an upgrade to a fundamentally new operating model — but I no longer think it’s unfixable. (Yes, I’m describing the thing I now spend my days on, so weight this accordingly.) Structural change is now possible, but the conditions that will allow for that have only recently arrived. The understandable sense of futility has come from three pretty daunting conditions.

First, political will has been chronically insufficient — for reasons that are rational, if frustrating. Investing in government’s plumbing doesn’t itself buy you anything with voters (though the impact of that better plumbing often does). Nobody cuts a ribbon on a hiring process. The benefits are diffuse and accrue over years (and the payoffs are often the dog that didn’t bark, grabbing you exactly zero headlines) so reform itself earns few points on the timeframes politicians operate on. The costs, on the other hand, arrive immediately, from folks whose cheese you’ve moved and whose power bases you have disrupted. And state capacity reforms need sustained follow-through across administrations, which is exactly what our politics is worst at providing. I’ve watched reform-minded leaders conclude, not unreasonably, that structural change means planting trees whose shade some successor will take credit for. So they hold a hearing on the symptoms they can point to, sidestepping the cause, and the operating model underneath stays put.

Second, the ideas haven’t been where they need to be. Too much reform effort has gone into working around broken systems instead of the harder slog of fixing them — work that produces value, but extends the life of a failing operating model. When more ambitious ideas do exist, they’re often not “policy-ready”: long on diagnosis, short on the implementation specifics and hands-on support a government leader needs to operationalize them. Some lack an evidence base because changing systems is one of the trickier things to measure. Many fail to understand the complex, adaptive nature of bureaucracies and offer naively deterministic solutions that only make things worse in the long run. And some smell too strongly of one faction or another to attract the broad coalition this work requires.

Third, and probably most importantly, there has been no shared agenda. People and organizations are working on pieces of this problem — civil service reform here, digital practices there, legislative modernization somewhere else — but there has been no common map of what needs doing, in what order, and how the pieces add up to a government that works. There’s plenty of precedent for the effectiveness of a shared agenda (and plenty of lessons to learn from past successes, especially about how to forge ideologically diverse coalitions). The marriage equality movement was made up of disparate organizations working a common state-by-state roadmap that turned an impossibility into settled law; the conservative legal movement built the Federalist Society’s talent pipeline and a common jurisprudential agenda. Transformation comes from fields with a plan, not projects with potential. State capacity has had a thousand worthy projects and no way to say what the field is trying to accomplish — so funders can’t see where their dollar fits in a larger and more ambitious framework, organizations can’t articulate how they will complement each other’s work, and public servants, including elected and appointed leaders, have few ways to find and enlist outside advocates and allies.

What’s changed

Each of these barriers is now starting to crumble, and is likely to crumble further as government faces oncoming disruptions: the contingent kind (pandemics, climate shocks, geopolitical crises), the political kind (DOGE, whatever the cost of its methods, made the brittleness of government’s operating model impossible to ignore and created openings for structural arguments that previously had no traction), and the structural kind, where AI towers above all others. These crises not only summon will, they also make the plumbing visible. But even absent major crisis, there is finally a cross-partisan convergence of people who recognize, for different reasons, that the system is producing outcomes nobody wants — and 36 governors’ races this year in which candidates of both parties are running on making government actually work. The political will is not yet sufficient. But for the first time in my career, the wind is at reform’s back instead of in its face.

The ideas are catching up too. We’ve gone from years of what I call “the recitation of the maladies” — constant decrying of how byzantine and ill-fitting the system is for the job we need it to do — to a new (but still nascent) wave of concrete, actionable proposals to fix procedural barriers like the PRA and to transform state and federal civil service systems. (The downpayment on this last big promise came out last month, and is likely too long for all but the most dedicated professionals to read. Thank God we have such dedicated professionals in Congressional offices and other think tanks.) These ideas span the key competencies that make up government’s operating model, and they span the “stack” (so to speak) from law and policy down to the day to day practice in bureaucracies that is the real test of whether government really works, like California legislators building feedback loops between the laws they pass and what those laws produce. And the field is learning to distinguish work that extends the life of the failing system from work that creates the conditions for a new one, like the FAFSA team parlaying a rescue into durable structural change, including shaping its own oversight environment and states simplifying programs so they need fewer people to administer and serve people better. AI accelerates this: when a small team can map a regulatory regime in a week or stress-test a policy against thousands of edge cases before it’s enacted, ideas mature faster, and the excuses for not acting on them get thinner.

And the shared agenda is finally being built. That’s what Recoding America³ (which I chair and advise) is focused on: building a coherent field around four competencies government must have — the right people, the right work, purpose-fit systems, and incentives for outcomes — and connecting and backing an ideologically diverse coalition of organizations whose efforts add up to more than the sum of their projects. Nan calls for making the problems we need to solve more legible. The organizations we help coordinate are doing this for their constituencies; we put those pieces together for a bigger picture. There’s much more we need to do there, but Nan’s call to action is well-received in government reform land.

Which brings me back to her architecture. The function Nan says the third wave needs most is independent allocators, organizations who pool capital and deploy it with judgment toward a common goal. The good news is that others have been seeding this idea for a while in a variety of domains, and there are many field catalysts and pooled funds who might meet Nan’s criteria for third wave ambition, energy, and absorptive capacity. Recoding America is a new independent allocator designed to move government into an operating model fit for the AI era. It likely wouldn’t exist if Kumar Garg and his colleagues at Renaissance Philanthropy hadn’t pushed for it to happen. Because they got us thinking about this play (which, to be honest, was really not on my bingo card for my post-book life), we’re far further ahead than we might have been in building pretty much what Nan is calling for, and others are building similar vehicles in different domains. Like others, we’ve identified far more work that needs doing than the money we’ve raised so far (~$40M) can fund, and we are actively spending that money. If Nan’s right that third wave philanthropists will want to trust in these vehicles to take big bets with Silicon Valley vibes, then it’s possible the train tracks being built from each side will meet in the middle.

Metrics blind spots

I promised to return to one of the many insights Nan offers that go beyond smart back of the envelope math. She correctly identifies another way that this third wave will need to distinguish itself from what’s gone before:

We should be conscious that we are entering this wave with an affinity for the measurement-oriented tools that defined Wave 2. These will be poor fit for some of the questions that will matter most in Wave 3. This isn’t an attack on those tools; they’re among the best things philanthropy has ever produced. But they weren’t designed for questions of civilizational flourishing, meaning, and what makes a life good – all of which Wave 3 will eventually need to grapple with. We’ll need to get comfortable broadening our decision making tools to include squishier instruments like taste and good judgement to have the best shot at answering these.

The thing Recoding America does — building state capacity— and the way we do it — developing and orchestrating a cross-ideological field — have both been hugely neglected over the past decade. The democracy field is a tiny fraction of overall philanthropy; within that small slice, government capacity gets just three cents of every dollar. I don’t have a number for field building, but Bridgespan has been vocal on this topic, saying that “Field-building efforts are one of the most valuable investments funders can make, but historically such efforts are the least funded.” A key reason for both of these is that state capacity lacks the legible metrics that attract first and second wave philanthropic dollars. Those tools are too often optimized to track the outputs of workarounds, not to solve underlying operating system failures, and ironically, they often lock organizations into driving a pre-determined measure in a certain direction just when they should be acting on what they’ve learned and pivoting to the deeper problem. (My take on mitigations for this dysfunction is here.) Direct service and domain-specific policy both pass the metrics test, so big money has gone there. When those programs fail to solve the problem, the impulse is to fund a different program rather than fix the operating model underneath that is quietly acting as the constraint. “Getting comfortable broadening our decision making tools” is good advice not just for the third wave, but for all philanthropy that wants to be part of meeting the disruptive moment we are walking into.

AI as an unlock

This third wave of philanthropy from AI founders is particularly well-suited to the task of fixing government because AI is itself part of what makes changing government now possible. It can tee up decades of accumulated procedural cruft for spring cleaning — yes, legislators and regulators still need to act on these right-sizing proposals, but they’ve never had the opportunity to in the past, because these dense tangles of law, policy, regulation, guidance, and practice were so impenetrable without the help of large language models. It can collapse the cost of software so that small government teams can build what once required nine-figure vendor contracts, enabling far more rapid test-and-learn cycles. It can provide our lawmakers speedier, more granular insights to inform policy responses. None of these are going to just happen on their own, of course, but isn’t that what philanthropically-funded education and advocacy is for?

Going upstream

Fixing government doesn’t have to come at the expense of the kinds of direct support that can save or improve lives immediately, like bednets that prevent malaria. In fact, part of Nan’s point is that even what have seemed like hugely scalable programs, ones for which the constraint has been simply money, could hit absorptive limits when this new wave of funding hits. This doesn’t mean these programs shouldn’t be aggressively funded, but they can’t be the only answer.

But there’s a more important reason they can’t be the only answer. If government is the institution we charge through the democratic process with (in the US at least) creating the conditions for life, liberty, and the pursuit of happiness, then public sector dysfunction is upstream of much of the human suffering those direct interventions seek to alleviate. As Abi Olvera points out, while bednets are definitely a great philanthropic investment, obsession with the math (lives saved/dollar) can blind people to opportunities that aren’t shaped that way. As she points out, there’s been a malaria vaccine since the 1990s, but the lives it could have saved were lost because of bureaucratic inertia, as clinical trials and agency reviews dragged on. “After Phase III trials confirmed it worked in 2015,” she writes, “the WHO required an additional multi-year pilot program due to meningitis cases in a few tested regions. Those cases turned out not to be caused by the vaccine, which most observers expected. No institution in the approval chain was required to weigh that six-year delay against the vaccine’s 13% reduction in child mortality.” If government dysfunction is seen as a fixed constraint, saving those lives is off the table.

There’s an old parable about villagers who hear a child drowning in a river, and pull him safely to shore — and then hear the next cry, and the next, and continue to pull children from the water. Finally, one of the rescuers breaks away from the group and heads up river. The others ask where she is going, as her rescuing skills are needed. “I’m going upstream to find out why they’re falling in,” she replies. There are versions of this story in which there is only one rescuer, and she has to decide whether to keep pulling the kids out of the river or go tackle the (literal) upstream problem, and that’s a much harder choice. But the group of villagers version is a better analogy for the moment we are entering, given the resources this philanthropy could bring. We can save the kids from drowning and try to stop them from being tossed into the river at the same time.

Nan asks what the third wave will do differently. Here’s what I’d add to her answer: it could be the first to actually try to fix government — not route around it, not tinker at its margins, but move it from the industrial-era model we inherited to one that can do what we need it to do in a world increasingly reshaped and sped up by AI.⁴

The reality is that we haven’t really tried — or at least, we haven’t tried this.⁵ What little philanthropy has been aimed at government capacity has mostly gone to vertical interventions: a high-skilled fellow parachuted into one agency, a special authority won for one team, a rescue squad that nurses a failing system back to “good enough.” Each does good in the moment. But they substitute for the broken system instead of fixing it, and they leave the system a little more complex each time, so the next fix starts from an even worse baseline. We’ve gotten very good at staffing around the civil service system instead of fixing it, and at winning carve-outs from burdensome procedures instead of reforming them. That’s the move the third wave could play: update the system rather than hack it, and orient a whole field toward that horizontal work instead of a thousand disconnected vertical projects. It’s the bet that’s never really been seriously placed, but suddenly there are not only new tools and new resources, but also new ambitions. Those holding them might heed what Stewart Brand said to an earlier era of newfound power: “We are as gods and might as well get good at it.”

Nan’s advice will help them get good at it. But gods who skip the hardest, most important problem — the one whose neglect quietly raises the price of everything else — aren’t there yet.

1. There are many who disagree with these estimates. The labs’ IPOs are not a done deal so they could be much lower. They could be higher. I have not checked her math and don’t intend to. I take it as worth exploring how any amount of new philanthropy should be spent. :)

2. This doesn’t account for the share of US philanthropic dollars spent overseas, for which I can’t seem to find a good estimate. Someone help me out here.

3. The organization has been called Recoding America Fund, but we realized “Fund” was confusing — yes, we have a fund, but we’re a lot more than that. So we’re dropping “Fund” from the name, and in a few weeks will have a much-improved website explaining what the organization does and how we do it.

4. For more on this, read my longer essay on the three horizons or the shorter version.

5. When I say we haven’t tried, it’s no knock on the many indefatigable champions for reform who’ve been fighting for attention for decades. I mean that philanthropy hasn’t really tried, in the sense that I mention above: funding levels incommensurate with the size of the problem and the lack of field coordination to help these efforts add up to more than the sum of their parts.