Originally published on LinkedIn, 2026-05-21.

Last Wednesday a documentary film put a name to something I’d chased for a decade in Silicon Valley. The ghost I could never quite catch.

I’m on the board of the San Francisco Foundation. About ten of us gathered in the board room to screen a new film called Ghost in the Machine. Most of the room was junior staff and middle managers, the people who usually sit along the walls during board meetings, not at the table. We pulled the shades on the Bay and watched.

The film traces artificial intelligence back to an uncomfortable root. It draws a line from the Victorian men who first tried to measure and rank human worth, through the founders of Silicon Valley, to the AI boom we’re living in now. The argument is that the whole project rests on one old idea: that some people are simply more deserving of wealth and power than others, and that you can measure who.

When it ended, no one spoke. The whole thing felt insurmountable. The room was set up panel style, three chairs up front facing rows of seats. It felt wrong. We didn’t need a sage on a stage. We needed to be in conversation with each other. So we pushed the chairs into a circle, and the program officer moderating asked us for our reactions.

I was one of two panelists, with Lili Gangas. She went first. I stared at the ceiling, trying to turn a deep feeling into a sentence, and what came out was Stamped from the Beginning. In Ibram Kendi’s deeply researched, 511-page sweeping history of racist ideas in America, he argues that racist ideas didn’t come first. The self-interest came first, and the ideas were built afterward to justify it. The hierarchy was the alibi.

The hierarchy in Silicon Valley presents itself as intellectual. But it grew out of the same old project: the attempt to prove that human worth can be ranked and measured. And it still does the same work. It keeps a certain kind of person at the top and lets them feel they have earned it.

I know that hierarchy from the inside. I have an engineering degree and two master’s degrees from Harvard. I worked at Bain, at Bridgespan, and at Google for a decade. My credentials were unquestioned, and they helped me climb. Before I was promoted to Director at Google, my boss told me a C-Suite executive had asked her where I’d gone to school. I earned that promotion on the work. But a decade out of grad school, the question still got asked. And two Harvard master’s degrees were the answer that satisfied it.

I say all of this because the same assumption now sits under almost every conversation about the coming wave of AI philanthropy. It’s under Nan Ransohoff’s smart, widely shared piece on the third wave of philanthropy. It’s under every room where this new money is being discussed.

The assumption is this: the people best equipped to solve our hardest problems are the people who can think about them most rigorously.

I don’t believe it.

It comes with a method. Rank the problems by expected value. Trust the calculation over the testimony of the people living closest to them. Treat sentiment and local loyalty as bias to be stripped out, and call the result objective. But choosing whose testimony counts as evidence, and whose counts as bias, is the oldest bias there is. I understand the appeal. I’ve spent my whole life inside it, and I know how seductive it is to believe a hard enough mind can reason its way to the right answer.

This isn’t an argument against intelligence. It’s an argument against worshipping it, and against the quiet belief that the people who have the most of it have earned the right to decide everything else.

Here is what that belief does, whether it means to or not. It takes moral authority away from the people who have actually suffered, and hands it to the people who can argue most fluently about suffering.

Those are not the same people.

Moral authority doesn’t come from a diploma. It comes from proximity. The person who has lived closest to a problem knows its contours in a way no model captures. They know what exclusion feels like. They know what it costs to be born on the wrong side of a line. That knowledge is not soft, and it is not a lesser form of expertise. It is the most reliable expertise we have about how oppression actually works.

I learned this in the work, not in a seminar.

Years ago at Google.org I made a $2 million grant to a Black-led movement against gun violence. They organized the people closest to the shootings around an approach the evidence backed. The grant wasn’t large enough to need executive approval, but the company asked me to get a senior vice president to sign off anyway. I still notice that detail. The work made the system want a higher signature.

The leaders of that work had a moral authority no credential confers. One of them put it simply: the people closest to the pain should be the ones leading the response. But when I sat with other funders working on criminal justice, some of whom came out of effective altruism, those same community leaders were heard but not centered. They were treated as people to be assessed, not people to be followed. The prime seat went to the funders and the analysts. The ones who actually knew were expected to defer.

It confused the leaders. For a while it confused me too. Then I saw it was the hierarchy doing what it always does.

The history is worth being precise about.

Modern statistics was built, in large part, by eugenicists. Francis Galton coined the word, and developed correlation and regression. Karl Pearson, who gave us the correlation coefficient, held a literal academic chair in eugenics. So did Ronald Fisher, whose work underpins most of modern statistics, and who was still defending innate racial hierarchy as late as 1950. This isn’t ancient history, and it isn’t fringe. In 2020, University College London stripped Galton’s and Pearson’s names from its buildings, and Cambridge took down a window honoring Fisher.

I’m not saying the math is tainted by the men who made it. Regression works. An idea is not false because its author was a bigot. My claim is different.

Eugenics was never only a set of tools. It was a project: ranking human beings on a single scale of measurable intelligence, then deciding who deserved to flourish. That project didn’t end. It changed its name.

It became the IQ test and the SAT. The SAT was designed by Carl Brigham, who sat on the advisory council of the American Eugenics Society and had just published a book on the superiority of the “Nordic race.” He recanted years later. The test was already built. It still shapes who gets through the doors of the academy. When we’re told that the people best equipped to run philanthropy are the ones who scored highest and reason most rigorously, we are being handed the most recent costume of a very old idea.

That is the pattern Kendi described. People build these hierarchies to justify their own place inside them, then call the result merit. The worship of measurable intellect is its newest form. It seats its own architects at the summit and tells them they belong there.

And it didn’t only shape who got in. It shaped what got built. William Shockley shared the Nobel Prize for the transistor and then founded the lab that brought silicon to the orchards south of San Francisco. His engineers left to start Fairchild, which became Intel and much of what followed. Gordon Moore, who cofounded Intel, said Shockley put the silicon in Silicon Valley. Shockley also spent his last decades insisting that Black Americans were genetically inferior, and proposing that people with low IQs be paid to be sterilized.

The reverence for artificial intelligence is that same instinct pushed to its conclusion: the dream of an intelligence even greater than the men who already sit, by default, at the top.

When you hold that much wealth and power, you need a story that makes it feel deserved. Intellect, dressed up as something neutral and measurable, is that story.

Nan Ransohoff sees part of this, and her piece does real work, forcing the sector to reckon with the sheer scale of the money coming. She is right that the measurement tools of the last wave, the ones effective altruism gave us, are a poor fit for the questions this wave has to face. But the talent frame she builds on top of them smuggles the old hierarchy right back in.

Look at who appears in the world she sketches, and who does not. Her builders are tech founders. Her capital allocators are philanthropic VCs, modeled on Sequoia. “Tech-caliber talent” is the standard, and she predicts these funders will be wary, by default, of the people who come out of traditional philanthropy.

Nowhere in that ecosystem, funders, allocators, builders, do the people closest to the problems appear, except as the problems themselves. They show up the way they always have, as case studies and quotes in a deck, assessed by the room instead of seated in it. A third wave built on that frame will hand the pen to the same people who have always held it.

Nan also hopes we don’t lose sight of human flourishing: aesthetics, civic life, moral imagination, the cultural conditions for a good life. I share that hope. So ask yourself honestly where human flourishing actually lives.

It is not in Silicon Valley. It is not in high intellectualism.

It’s in cultures that never stopped valuing community, rest, nourishment, and connection to the land and to the sacred. It’s around a campfire. It’s in the things our most intellectual spaces have been quietly starving for. The academy was never built to credential the people who kept that knowledge alive.

So I’m not asking us to abandon intellect. I’m asking us to stop centering it. Not in who accumulates this capital, and not in who gets to distribute it.

The people we should be looking for are bridgers. People who carry the lived experience of communities that have suffered, and who can also navigate the academies and boardrooms well enough to move real money. People who can sit in both rooms and refuse to let either one erase the other.

That is harder to find than a brilliant analyst. It’s also the only thing that will keep this third wave from intellectualizing, one more time, on behalf of people it never bothered to include.

That night at the foundation, the people in the circle were not the ones who usually hold the pen. The circle was the right shape for the conversation. It may be the right shape for the money too.

The money is coming. The real question was never whether we’re smart enough to deploy it.

It’s whether we’ll trust the people who actually know.

Originally published on LinkedIn, 2026-05-13.

Vu Le’s essay on the harms of AI in the social sector dropped yesterday. He named real harms. I spent the day digging into the underlying research. The strongest claims hold up.

Our sector hasn’t been racing toward AI tools. The opposite. Most of us have been hesitating in the library while the music bumps in the next room. Vu’s piece arrived right as some of us were finally walking toward the door. It’s going to convince a lot of those people to sit back down.

Underneath the post is a different conversation. In the comments. In the DMs.

The harms are real. Refusing won’t fix them. Using carefully won’t fix them. Government isn’t coming in time. So what am I supposed to do?

I directed $700 million in philanthropy at Google.org. Last September I was on the other side, pulling retirement money to keep what I’m building with AI alive. Both seats taught me the same thing about where our power actually lives.

It’s not in the answer we keep grabbing for.

We’re asking the wrong question

The sector has been stuck on whether we use AI.

Frame it that way, and we’re consumers. Opt in. Opt out. Use it carefully. The companies that build these models will ship the next one whether we participate or not.

Our power is larger than that. What gets built, by whom, for whom, accountable to which mission, is still being decided. We help decide it. With what we fund. With who we name. With who we back.

Whether to use it is the wrong question. The right question is who we back to build the thing that comes after.

What Vu got right

His strongest claim is one we should read and remember.

Last year, researchers at Stanford and the University of Chicago fed leading AI models the same questions written two ways. Once in standard English. Once in Black English. Nothing else changed.

The models recommended more criminal convictions when asked to judge the speakers of Black English. More death sentences. Lower-status jobs. The researchers then trained the models to be more polite. The models stopped saying racist things out loud. They kept making the same decisions.

That’s a peer-reviewed result. If you build, fund, or deploy AI in this sector, you need to take heed. “The polite tone these models have been taught isn’t evidence the bias is gone.”

Vu’s other strongest claims rest on similar evidence. Court records. Peer-reviewed surveys. FEC filings.

The workers in Nairobi paid pennies to watch the worst content humans have ever made.

The surveillance. The political alignments of certain CEOs.

This isn’t moral panic. It’s the paper trail.

What Vu leaves out

He also leaves things out. They matter. They’re the case for what’s possible.

A purpose-built therapy chatbot, designed with clinicians, cut depression symptoms by half in a randomized trial.

AI cuts professional writing time by forty percent. It does the most for the workers with the least training, not the experts.

Blind, deaf, dyslexic, and motor-disabled users have built workflows on these tools that didn’t exist three years ago.

None of this refutes Vu. It shows a different thing. AI built for the work, by people accountable to the work, can do what general-purpose AI can’t. Refusing all of it gives up that ground.

That’s the case for backing builders.

The three answers we reach for

Most of us reach for one of three answers. Each feels like agency, but none of them changes what gets built.

Refuse it. A personal stand has its own integrity. Bearing witness without complicity matters. Witness alone doesn’t build something different. The companies don’t need your usage to ship the next model.

Use it carefully. Necessary. Not enough. Careful use makes a single workflow a little better. It doesn’t change who decides what gets built next.

Wait for government. Keep pressing, especially at the state and city level. Congress is locked. The industry has put more than a hundred million dollars into political efforts designed to weaken or preempt state AI safety laws. The federal action we need isn’t coming in time. Policy matters. Waiting isn’t a strategy.

So the question is real. What’s left?

Where our power actually lives

What’s left is the people close to the work.

The AI architecture is being built by the frontier labs. The application, what tools serve what mission, is still ours.

There’s a layer of AI being built right now by people who came up inside this sector. Some are running pilots inside nonprofits. Some are bootstrapping on personal savings. Some are running open-source experiments out of university labs. Most are invisible.

That’s the layer we can change.

Refusing doesn’t build it. Using carefully doesn’t build it. Waiting doesn’t build it.

People we back build it.

Three things to do right now

Three concrete moves. The dollars aren’t equivalent, but the importance is.

1) If you sit in a funding seat. Write at least one check this year to an organization building something socially impactful with AI. A hundred thousand to a quarter million dollars, structured so success recycles the capital and failure still counts as a mission-aligned grant. The instruments exist. Your foundation counsel knows them.

Foundations move a hundred and ten billion dollars a year. About a penny on the dollar comes back as recoverable capital. Almost none of that penny reaches the kind of AI being built by startups. What’s missing is the practice.

Before you write the check, hand them Vu’s essay. Ask which harms their work makes worse, which ones it doesn’t, and how they know. A builder who flinches at that question isn’t ready to build for our sector.

(The longer version of this argument is in I Gave Away $700 Million. Then I Tried to Raise $350,000. This is the structural ask underneath it.)

2) If you lead inside a nonprofit. Pick one workflow your team would build differently if you owned the tool. Name it. Tell a funder what it is. The next time someone asks how your team is using AI, flip the question. “We know the tool our work needs. Will you help us build it, test it, or buy it from someone whose mission lines up with ours?”

3) If neither of those describes you. Make a builder visible. Most organizations working on this die for one reason. Nobody can find them.

Name one of these orgs publicly in the next thirty days. Hire them for a small piece of work. Subscribe to what they ship. Send their name to a funder you know. Repost what they write. Visibility is its own capital. It’s also the move that doesn’t require a checkbook.

Enough of these, by enough people, changes what becomes normal.

What this won’t do

Backing builders doesn’t stop the frontier labs from shipping what they’re shipping.

It doesn’t replace policy. We still need state attorneys general. State legislatures. Federal action when it finally comes. Keep the pressure on.

It doesn’t make the harms disappear.

And it doesn’t put the public in charge of AI.

That’s the deepest critique of this argument. The people deciding what gets built are still a small group choosing for the everyone else. Funders. Founders. Networked sector folks. The asking is happening in the rooms we already sit in, not in the rooms where the cost is being paid.

A question that belongs to the public doesn’t have an honorable private answer. The fight for AI that’s democratically governed, owned and steered by the people whose lives it’s reshaping, is the bigger one. It’s also the right one.

Public ownership stakes on AI infrastructure are on the table for anyone with the nerve to put them there. That work matters. We should be loud about it.

It’s also a multi-decade project, absent a catastrophe that shifts public opinion fast. In the meantime, the question is what we do with the agency we already have.

Philanthropy moves more than enough capital to measurably move the needle on what gets built in the next four quarters. And the four after that. And the four after that.

Not as a substitute for democratic governance. As the work of the meantime.

What it does is seed a different ecosystem. Tools built by people who understand the work. Accountable to missions that aren’t solely focused on shareholder returns. It moves the ground under our feet, even a little, toward the kind of AI this sector would actually trust.

That is not nothing.

What’s left to do

The next time a colleague forwards Vu’s essay and asks what to do, the honest answer is shorter than the conversation we’ve been having here.

The music isn’t stopping. The architecture is being built. The next model is shipping whether we participate or not.

Refusing won’t change that. Using carefully won’t change that. Waiting won’t change that.

The people who will change it are already in your network. Most of them aren’t visible enough yet.

Find one. Back them. Tell the rest of us who they are.

Who’s building the AI this sector would actually trust?

I’ll start. We built the Kindora Claude Connector last month. It lets people search more than a hundred thousand open grants from inside any AI assistant. We gave it away free.

People keep asking why we didn’t charge for it. The honest answer is that’s what AI built for our work should do. Save people time they can reinvest in relationships. Make their work stronger. Delight them. We believe generosity will come back one way or another.

One example. Not the only one. Not even the best one. The one I know best.

Share yours.

Originally published on LinkedIn, 2026-04-28.

September 29, 2025. I was at my desk in our backyard cottage in Oakland, late, Sally Steele in the house doing her own work. Fifteen clicks deep into Vanguard’s withdrawal screens. About to pull $100,000 from our retirement.

It was terrifying. This was our retirement.

2025 had been a brutal year for nonprofits, and a brutal year for me. I’d spent it scratching for consulting clients, on unemployment for the first time in my life. One client reneged on a success fee the moment it looked like it might trigger. Another had me working for six weeks before I walked away from a contract their lawyers wouldn’t finalize.

Sally was running Outdoorithm, the social enterprise we co-founded to get more families outdoors, and she wasn’t paying herself either. Four daughters, ages 4 to 16, all sharing one bathroom in a small craftsman. Money going out faster than it came in.

I was already $50,000 in on Kindora, the AI fundraising platform my friend Karibu Nyaggah and I had launched months earlier. It was working. Fifty nonprofits signed up off our first LinkedIn post. A VC-backed company offered to acquire us. A private equity firm called. Another VC asked if we were raising. The private sector was happy to give us money.

What I wanted was $350,000 from the sector I’d spent my career building. At Google.org I’d directed a team that gave away over $100 million a year. We rarely made grants this small. I knew the people running the major foundations on a first-name basis.

That check wasn’t going to come from any of them.

I clicked transfer. Closed the laptop. Some of the money pressure rolled off. The pit in my stomach didn’t. We were spending our acorns while everyone else around us was stacking them.

The phone calls

I’d already made every call I knew to make.

Right after we launched, I started reaching across my philanthropy network. Friends I’d worked alongside for years. People I deeply respect, who had seen my work up close. The pitch was simple. We bootstrapped this with our own money. The private sector is circling. We’d rather take investment from mission-aligned funders.

One conversation stays with me. A leader at a major philanthropy told me he’d personally invest his own money. He said people in philanthropy rarely get the chance to participate in deals like this. I appreciated the vote of confidence. But I wasn’t looking for personal angel checks yet. I was looking for institutional capital. The kind of patient, mission-aligned investment his organization existed to deploy. That path didn’t open.

The pattern was consistent and well-intentioned. They pointed me to earned revenue and ended the conversation.

The tools to invest in mission-driven startups exist, they’re just rarely used. It’s not that funders don’t care, it’s that precedent is thin and teams are reticent to introduce more risk and complexity. For most funders, the playbook hasn’t been written yet.

The VC trap

While the philanthropy calls were going nowhere, I was studying venture capital. Hard.

September 2025. Same month I was making those calls. I was at Responsible Innovation Labs’ EDGES gathering in San Francisco. Roy Bahat, managing director of Bloomberg Beta, was on stage. During Q&A I asked whether he’d raise VC if he jumped back into founder mode.

His answer didn’t take long.

“The only good reason to raise VC is you have no other choice.”

Six months later I heard the long version of why. In March, at SXSW, I sat in a founders house and listened to Dug Song. He co-founded Duo Security, Michigan’s first tech unicorn, acquired by Cisco for roughly $2.35 billion. Before Duo, he co-founded Arbor Networks. That story went differently.

Arbor raised over $120 million in VC. A $100 million acquisition offer came in year three, but the capital structure made it impossible for the founders to take it. The company was eventually acquired at terms that gutted the founders. Song made less than the employees of a company they’d been forced to acquire. He had to leave the company he started.

“Every dollar you raise creates a valuation hurdle you must clear,” he told us. Raise $10 million when you only need $2 million, and you’ve pushed your milestones so far out that missing them kills your optionality.

Bahat had named the verdict. Song had shown me the math.

The no-man’s land

Here’s the conundrum for founders building technology for the social sector.

You can take VC. For the right company, it’s rocket fuel. But most businesses don’t need rocket fuel. They need regular gas. They need to drive somewhere, serve customers well, grow at a sustainable pace. For tools meant to be accessible and durable, signing up for rocket fuel means signing up to become a billion-dollar company or go bust.

Many funds today want 10x year-over-year growth. If you’re not on that curve, you’re not attractive. And 10x growth means you can’t maintain culture or consistent staff. In the social sector, mission comes through culture, through relationships, through people. Try growing that by 10x every year.

You can become a nonprofit. But then you can’t offer equity to attract talent. You can’t share in the upside if you succeed. And philanthropy has a poor track record of sustaining technology investments. I watched this play out at Google over and over. Nonprofits building technology that never reached the scale or quality of for-profit alternatives.

You can bootstrap. But you need some capital to get started. The people closest to the problems worth solving are often the least likely to have wealthy networks of angels willing to write checks. Or retirement accounts to drain.

When I asked Katie Boland at Delta Fund about alternative structures, she put it plainly. “All of the above are good, but not good for everyone.” There’s no central clearinghouse. No established playbook for founders building for-profit, social-purpose technology.

What the data actually shows

Foundations move $110 billion a year in grants. They deploy $1.4 billion a year in program-related investments. About one cent on the dollar in mission-aligned, repayable capital designed for companies like ours.

I know this because we built it. Kindora’s database covers three years of IRS filings. More than 170,000 foundations. Nearly 7 million grants. Every program-related investment U.S. private foundations have reported. The pattern isn’t subtle.

That penny is heavily skewed. The biggest foundations (Gates, Ford, MacArthur) make most of these investments. Of 5,363 U.S. foundations holding more than $100 million in assets, only 193 made a PRI in the last three years. That’s 3.6%. More than 96% of large foundations aren’t using a tool that’s been legally available for decades.

Where the money does flow, it goes to housing and community lenders, not technology. Across three years of filings, we identified 26 direct PRIs to for-profit technology, software, or AI companies. About $50 million total, out of $4.2 billion in PRI volume over that period.

Three checks account for $43 million of that $50 million. Two PRIs from Steele Foundation for Hope (no relation) to Sibel Health and Zenysis Technologies, $38 million combined. One from the Gates Foundation to Osmo Labs, $5 million. Strip those three out and you’re looking at $7 million scattered across 23 companies, most under $500,000.

Steele Foundation for Hope is interesting. A New Hampshire family foundation founded by Stuart and Suzanne Steele in 2022, explicitly built around deploying PRIs to for-profit companies as its primary mode of giving. When a foundation is structured this way, it can write $25 million checks. The fact that almost no other foundation in the country operates this way is the story.

The Richard King Mellon Foundation in Pittsburgh is the other standout. They make 10+ investments a year in the $150,000 to $500,000 range, directly to early-stage tech companies. Marinus Analytics, an AI anti-trafficking startup. Peoples Energy Analytics, an energy bill assistance platform.

Then we ran one more search. We looked through every PRI to a for-profit tech company in our database for one specific thing: companies building technology for the nonprofit and philanthropy sector itself. Grant management software. Donor databases. Fundraising platforms.

We found one. A $250,000 investment from the James Lee Sorenson Family Foundation in Chario, an early-stage giving tool. One adjacent match, Brighthive ($50,000 from Lumina), builds data infrastructure used by nonprofits, government, and education providers. That’s it.

A sector that manages $5.7 trillion in assets and distributes $110 billion in grants every year is not using its own investment tools to fund the technology that powers its own work.

The void didn’t stay empty. VC filled it. Every for-profit software company that scaled in philanthropy took venture capital or private equity. Instrumentl. Fluxx. Submittable. Bonterra. Classy. Daffy. The tools got built for the customers VC cares about, not the 85% of nonprofits priced out at $300 to $500 a month. That’s the bill that comes due when philanthropy doesn’t fund its own infrastructure.

I’m not pointing fingers. In ten years of grantmaking at Google.org, I only ever made a handful of ER grants. Zero PRIs. I made the case to leadership for building a PRI portfolio. It didn’t happen. Even with the right experience and a strong conviction about why it mattered, I couldn’t get one off the ground.

Why AI changes everything

Here’s what gives me hope.

My wife and Outdoorithm co-founder Sally studied Foreign Affairs and French. She’s pushing code now. During Black History Month, she decided to add heritage badges to campgrounds on the platform that had significance in the Black experience. She wrote the API calls. Tagged the data. Created the icons. Deployed it herself. A year ago that would have required a developer.

What used to take a 50-person engineering team can now be built by two or three people with domain expertise and AI tools. We built Kindora’s core product in a week. The cost of creating world-class software has collapsed.

This changes the math. With a lean team, you reach profitability faster. You grow through revenue instead of fundraising. A $10 million to $50 million company becomes viable and desirable, not a consolation prize.

The technical wall fell. The capital moat didn’t.

Revenue-funded doesn’t mean zero-capital. You still need a bridge. By the time we closed our first outside check, we were $80,000 in. There are server costs. AI inference costs. And eventually you need to pay people. The bridge from $0 to sustainable revenue is 12 to 24 months. That bridge capital is exactly what the social sector’s capital stack doesn’t provide.

What could work

The instruments exist. Many people in the sector don’t understand how they differ. Three are worth knowing.

Recoverable grants. A founder gets, say, $100,000 that gets paid back only if the company hits specific revenue thresholds. If it doesn’t work out, the grant is forgiven. The money that comes back funds the next cohort. Echoing Green pioneered this. It’s the closest thing the social sector has to “believe-in-you money.” But Echoing Green can do it because of how it’s structured legally. Most foundations can’t copy the model directly.

Expenditure responsibility (ER) grants. A foundation can grant directly to a for-profit company. The catch: every dollar must go to a specified charitable activity, and the foundation has to monitor compliance. The McGovern Foundation has made this work at scale, with 70+ recent ER grants to for-profit technology and data companies. Climate Policy Radar got $500,000 for AI-powered legal research. Open Function Group got $400,000 for open-source data integration. The legal path exists. McGovern walked it. Few followed.

Program-related investments (PRIs). This is the tool that should be doing the heavy lifting. A PRI is an investment, not a grant, where the primary purpose is charitable. It counts toward the foundation’s required annual distribution. It can be structured as equity, a loan, convertible debt, or revenue-based financing. If the company succeeds, the foundation gets its money back. If it doesn’t, the foundation writes off the investment. Same outcome as a grant, with upside potential.

When most people say “PRI,” they picture a below-market loan to a housing developer. But a PRI can also be a forgivable loan: zero percent interest, forgiven if revenue triggers aren’t met. That’s exactly what an early-stage founder needs. Sorenson Impact runs an active portfolio of 35 PRIs totaling $5.2 million in education, health, and financial inclusion. Back in 2017, Kresge made an early PRI to Fluxx, a grants management platform built for the philanthropy sector itself. These foundations figured it out.

Donor-advised funds can deploy these instruments too. Americans hold $250 billion in DAFs. Most of it isn’t moving fast enough.

Camelback Ventures provided Kindora with a $50,000 SAFE note that was, frankly, a lifeline. Patient capital from a mission-aligned investor who would be happy if we had massive impact even if the financial return was modest. That’s the spirit of what’s needed. We just need it at a scale that doesn’t require founders to get lucky.

The model that works for companies like ours is straightforward: revenue-funded in the long-term, mission-bridged in the near-term. The free tier of our platform, serving nonprofits who can’t afford existing tools, is a genuine charitable project. Fundable through any of the three tools above. The paid products are a real business. The bridge capital buys the runway from zero to sustainable revenue. If the company succeeds, the capital recycles. If it doesn’t, the charitable work still happened.

By April 2026, Kindora was serving over 700 organizations, with an Anthropic partnership that offered free open grant searches in Claude. Revenue was picking up. We weren’t asking to start. We were asking for the bridge.

The call

At The Light House at SXSW, hosted by Omidyar’s Tech We Want team, Jessica Norwood gave me a name for what was missing. “Believe-in-you money.” Reparative capital that says, “I think you should have a shot at trying.”

That’s what’s missing from the social sector’s capital stack. Not because people don’t believe in founders. Because this is a muscle the sector has rarely exercised. Backing mission-driven founders, at the early stage, with patient capital, in for-profit structures.

This isn’t just about Kindora. It’s about hundreds of founders building social-purpose technology right now, facing the same impossible choice. People closest to the problems but cut off from the patient capital to solve them.

The sector’s response so far has been more reports and more convenings. Long guides on how to fund AI. Coalitions still debating what to fund and how. Meanwhile, founders are draining retirements.

I understand the reflex because I lived inside it. When a decision felt risky, the safest move was another framework. Another report. Another convening. Those tools can clarify a decision. But too often, they become a way to avoid one.

Stop convening. Start writing checks.

Set aside 5% of your annual grants budget for forgivable PRIs or ER grants. As a start. $100,000 to $500,000 per check to founders building social-purpose technology who need the bridge from $0 to revenue. Capital that comes back recycles into the next cohort. The rest is a grant, like the other 99% of foundation funding already is.

Don’t overthink it. Define the impact you want to see. Set up the vehicle. Get the money out the door. Learn. Repeat.

Publish the terms when you deploy. Diligence memo, term sheet, repayment triggers. Make the next foundation’s first program easier than yours was.

You don’t need investment managers squeezing every dollar of return. This is just a grant that might recycle.

We’re living through a once-in-a-generation shift. The cost of building has collapsed by an order of magnitude in 24 months. A platform that used to cost millions and a team of eight now costs $100,000 and a founder with the right tools. Domain experts who were locked out of building for their entire careers can finally create what their communities need.

Foundations move $110 billion in grants every year. About one penny on the dollar gets deployed as a PRI. The cost of leaving that gap open is that we keep sending the founders closest to the hardest problems into a capital market designed for someone else.

Someone has to go first.

If you sit inside a foundation, who deserves a shot at trying?

Originally published on LinkedIn, 2026-02-17.

Think back to the summer of 2020.

Not to COVID. To what happened inside of it.

We’d been watching Black people die on camera for years. Eric Garner. Tamir. Alton. Philando. Each one a wave of outrage that crested and receded. Protests, then silence. Hashtags, then nothing. Like school shootings. Unbearable for a week, then absorbed into the background noise of American life. You’d feel it in your chest, and then you’d go back to work.

Then George Floyd.

Same violence. Same cameras. But something tipped. Maybe it was the length of that video. Maybe it was COVID locking everyone in their homes with nowhere to look away. Maybe it was nine minutes and twenty-nine seconds, long enough that you couldn’t tell yourself this was a split-second mistake.

In Oakland, Sally Steele and I walked our four daughters over to Park Boulevard. Hundreds of families were already there. We knelt together for eight minutes and forty-six seconds — the number the country was counting then. My oldest held my youngest. Nobody said a word. You could hear the birds.

Then everyone stood up. And the country erupted.

Not because the facts changed. Because something tipped, and people who’d been on the sidelines their whole lives stepped into the streets. Researchers at Harvard estimated that 15 to 26 million Americans protested that summer, quite possibly the largest protest movement in the country’s history. It wasn’t organized by one person or one institution. It was a force that had been building quietly for years until the moment it became unstoppable.

I’d spent the five years before that summer directing over $100 million in racial justice funding at Google, building partnerships with organizations like the Equal Justice Initiative after the Charleston church shooting, trying to meet each crisis with resources and infrastructure. I know what a tipping point feels like from the inside. I’m not comparing the stakes. I’m naming the pattern: when the people who’ve been sidelined step into their power, systems change. And I think we’re in the early seconds of another one.

Not about racial justice directly. About who gets to build things. About who gets to create the tools that shape how we live. About a wall that has separated the people who understand problems from the people who build solutions for all of recorded history. The same question that drove those protests — who holds power, who gets to wield it, whose lives get shaped by decisions they had no hand in making — is now playing out in technology.

That wall just fell.

An essay called “Something Big Is Happening” went viral earlier this month. A tech founder named Matt Shumer wrote it as a letter to his friends and family. He compared this moment to February 2020, right before COVID changed everything. He said AI had quietly gotten so good that it could do his job, writing software, better than he could. He told people to start using AI or get left behind.

He’s right that something big is happening. But he’s telling a story about what AI takes away.

I need to tell you what AI gives. And I need to stop being polite about it.

Here’s my confession

I’ve been moving through my world, through nonprofit boardrooms and funder convenings and Oakland basement meetups, and dialing myself down.

I spend my days building AI-powered platforms. Coding until one or two in the morning. Shipping software over weekends. Building things in hours that would have required entire engineering teams two years ago. It feels like being in a nightclub at peak hour. Relentless. Electric.

Then I walk into social impact spaces, and it’s like stumbling out of the nightclub into a library. My ears are still ringing. The beats still reverberating in my body. I’m bobbing to the music in my head as I walk through the doors. It’s quiet. Folks are looking at me like I’m being disruptive. But I’m on a high. I start talking about how alive things are in the AI building space. My enthusiasm is real. People can see it. But they’re not convinced. Not convinced they’re invited to the party. Not convinced the tools are as capable as I say they are. They tried them a few months ago. Good but not great. Made mistakes. They’re not coders, so they don’t think they can code with them. You need a real developer.

So I dial it back. Match the room. Speak carefully.

“AI has some compelling applications for our sector.”

“After you understand the potential harms, it’s worth exploring.”

“Start with low-risk personal use cases.”

That’s the polite version. The version that doesn’t make people uncomfortable. When I talk at full volume about what I’ve seen, people look at me like I’m being naive. Like I don’t understand the real constraints. Like I’ve been spending too much time in Silicon Valley and not enough time in the real world.

But I have four daughters. I live in Oakland. I spent nearly a decade inside Google directing $698 million in philanthropy to communities that were promised innovation would help them and then watched it mostly help the innovators instead. I know the real constraints. I’ve lived inside them.

And I’m telling you, from inside those constraints, that what is happening right now is not incremental. It’s not hype. And the people I care about most are going to miss it if someone doesn’t say this plainly.

So let me say it plainly.

Over a bag of fries

Shumer’s moment of truth came on February 5th, 2026. Two AI labs released new models and he realized AI could now do his entire job better than he could.

My moment started three years earlier. Over a bag of fries.

December 2022. Sally, the girls and I are driving home from Yosemite with michael mcbride and his family. We’ve been dreaming about creating something to get more families like ours, from urban cities, into national parks and forests that feel like they belong to everyone but somehow don’t feel like they belong to us.

On the drive, we try to name this thing. Rooted Outside. Nature Rhythms. Every good name is taken.

ChatGPT had been released to the public a few weeks earlier. We open the app and tell it what we’re trying to build. At first it comes up with similar names as us. All taken. We stop at a Five Guys to grab lunch and keep prompting over a huge bag of fries and burgers while the kids climb over each other in the booth.

We ask it to combine two words into a new word, one that’s not in the dictionary, that captures the spirit of what we’re building.

Outdoor + Algorithm = Outdoorithm. With a double meaning: finding your rhythm in the outdoors.

We know immediately. That’s it. A Google search confirms zero results. We’ve just created something brand new in the world. We buy the domain on Squarespace before we get back on the highway. A week later, DALL-E creates our logo.

That was a name. A fun trick. A parlor game with a new tool. But it planted something.

Six months later, June 2023, I want to help our camping families understand what weather to expect at any campground, any time of year. I ask ChatGPT how. It recommends writing a Python script that pulls historical weather data from GPS coordinates.

I’d written C++ in engineering school twenty years earlier and built a basic website in HTML. But I’d never touched Python. Never built a real web application. Never written anything beyond homework assignments that ran in a terminal.

I ask ChatGPT how to install VS Code on my computer. I give myself one night.

Two hours of copying and pasting code from ChatGPT into this tool I haven’t touched since college-era programming. I run the script. It spits out a file. I open it.

A perfect spreadsheet. Fifty-two weeks in rows. Average high temperature, low temperature, and chance of precipitation at Pfeiffer Big Sur State Park campground.

I jump out of my chair.

It’s midnight. The house is quiet. Sally’s working in the other room. The girls are asleep. And I’m standing in our Oakland living room staring at a screen, realizing that something fundamental just shifted.

That was June 2023. I haven’t stopped building since.

The timeline nobody’s talking about

Shumer lays out the AI progression from the tech world. Here’s a different one.

2023–2024. I teach myself to build with AI, one project at a time. Copying code from ChatGPT, pasting it into VS Code, fixing what breaks, learning as I go. Interactive campground maps. Weather widgets. Backend data queries. Custom features for our community.

September 2024. After nearly a decade at Google.org, my role is eliminated. One meeting and an unceremonious exit.

Fall 2024. I discover Cursor, an AI coding environment that can read my entire codebase. Before this, I’d been copying and pasting code snippets into ChatGPT without the full context. As someone just learning to code with AI as a pair programmer, that was the hardest part — figuring out which pieces of my script were relevant for each new feature or bug fix. Cursor changed everything. It could see all of it. I use it to fully build out Outdoorithm’s campground map: thousands of public campgrounds from a patchwork of county, state, and federal agencies, color-coded by sentiment, with information popups when you click the icons. The code for that map was too long to fit in a single context window before. Now I can build it in days.

February 2025. 11 PM. Our small Oakland craftsman home. Sally is on the couch planning our nonprofit’s winter camping trip to Muir Woods. Outdoorithm Collective’s bank account: $33,000. The heaters we need for families camping in 45-degree weather: $9,480. I’m staring at 3,000 funder “matches” from a platform that costs $3,600 a year. Matches that would take a full-time person a year to sort through.

I pop open Cursor and build an AI system that thinks like a program officer instead of a search engine. I know how program officers think because I spent a decade as one. The AI evaluates each funder the way a real person would: reading about the organization, assessing fit intuitively, deciding whether to dig deeper or pass.

First pass: 90% of false positives eliminated. Three thousand matches become three hundred. Then one hundred fifty, stack-ranked by quality.

At the top: the George Family Foundation. Bill George was my professor at Harvard Business School. I’d been his student, his fellow, and I’d never connected his foundation’s focus on whole-person wellbeing to our outdoor equity work. The AI found a strategic match I missed for years. Within twenty-four hours, we have a meeting with their Executive Director.

March 2025. My friend Karibu Nyaggah, a bond forged twenty years earlier at Harvard, Google colleague, now at Meta, calls me from his kitchen island in Half Moon Bay. Three hours later, we’re co-founders. We call it Proxi at first, later Kindora. In April, we hire a senior full-stack developer to help take our prototype into production. $6,500 a month, out of our own pockets. At that point, AI models couldn’t quite take you all the way to production-grade software. You still needed a professional developer to close the gap. Progress is steady but slow.

August 2025. We incorporate Kindora as a Public Benefit Corporation, launch our beta, and I start using Claude Code, Anthropic’s command-line coding tool. Another leap. I prompt it in plain language from my terminal and it builds across my entire codebase. The first weekend, I ship an analytics dashboard, email automation for the flood of beta users, a feedback system, and support ticketing. Part-time. That’s work that would have taken our developer multiple weeks. Same team, ten times the output. Not because we worked harder, but because the tools had changed underneath us.

December 2025. We end our developer contract. Three years of building with AI had converged with the release of Claude’s Opus 4.5 model in November. Anthropic’s own engineers were building entire codebases with Claude. I found I could do the same. The gap that required a professional developer nine months earlier had closed completely.

By early 2026. Outdoorithm and Kindora are both production platforms serving hundreds of organizations — campground discovery, AI trip planning, funder matching across 175,000 foundations, grant intelligence, federal grants pipeline. I built both. I also build AI products for clients through my consulting practice.

Last week. I built a virtual voice AI assistant that simulates a funder conversation so nonprofits can practice their pitch before a real meeting. Four prompts. One night. Working prototype.

I’m 43 years old. I have four daughters, ages 4 to 16. I have degrees in chemical engineering, business, and policy. Three years ago, I had never written a Python script or built a web application. I now run technology ventures, ship production software, and build AI tools that serve hundreds of organizations.

If you think that story is about me being uniquely talented, you’ve missed the entire point.

The wall between knowing and building

Shumer’s essay tells a displacement story. AI is replacing human work. His job got automated. Yours is next. Learn AI or get left behind.

I’m living the other side of that story.

For as long as anyone can remember, there’s been a wall between the people who understand problems and the people who can build solutions. On one side: the social worker who’s watched families fall through intake cracks for fifteen years. The teacher who knows exactly why her students are struggling. The environmental justice organizer who sees patterns in pollution data that researchers miss. The restaurant owner on Foothill Boulevard in Deep East Oakland who can barely pay her electric bill and sits in a breakout room at the mayor’s economic forum asking, “Where’s that $100 million?” because the investment that was supposed to revitalize her corridor never reached her.

On the other side of that wall: the engineers. The people who could build software, tools, and systems but who’ve never met the families, never walked the corridors, never sat across from someone choosing between paying PG&E and buying groceries.

That wall is gone.

Shumer proves it from one direction. AI can now do what developers do. I’m proving it from the other. People who were never developers can now build. Both of those things are true. But only one of them changes who holds power.

When everyone can build, the differentiator shifts. It’s no longer about technical execution. It’s about knowing which problems matter and what good solutions look like. Call it taste. Not taste as aesthetics. Taste as the deep, earned understanding that comes from fifteen years in the field, from proximity to real communities, from doing the work.

Here’s what taste looks like. Outdoorithm, the camping platform Sally and I built, has a feature called The Green Book. Named after the historic Negro Motorist Green Book that helped Black travelers find safe places during Jim Crow. Our version uses AI to analyze 1.85 million campground reviews for discrimination, safety concerns, and community atmosphere, then flags which campgrounds are welcoming for families of color, LGBTQ+ travelers, and solo women. No engineering team at a camping startup would think to build that. I built it because I’ve been the dad loading four daughters into the car wondering if the campground is going to feel like it belongs to us.

That’s what taste is. The lived experience that tells you what to build when the tools finally let you build it.

For decades, Silicon Valley’s advantage was that they knew how to build. The domain experts knew what needed to exist but couldn’t create it. That asymmetry kept power exactly where it was.

It’s collapsing. Right now.

“You mean Claude can build that for you”

Last week I was at the Camelback Ventures accelerator with a cohort of founders building in education and social impact. We did a “reciprocity ring” exercise where each person wrote down what support they need and what they can offer.

Brianna Baker, the founder of Justice for Black Girls, wrote that she needs help building a tech platform to connect her research fellows with the seventy-five members of their Academic Network. Scheduling, automation, follow-up emails, connections between aligned research.

I tapped her at the table next to mine and told her I can build it.

Another fellow at the table chuckled. “You mean Claude can build that for you.”

As if that made it less real. As if my ability to understand what Brianna actually needs, to translate her vision into the right architecture, to exercise judgment about what matters, wasn’t the hard part. As if the only “real” way to build software is to type every character yourself.

That sentiment is everywhere in my spaces. And it’s leaving us behind.

EdTech founders at the same program told me their software needs to be airtight for school districts. No security risks, no authentication errors. They think AI-generated code can’t get there. I understand the instinct. Six months ago, they may have been right. But right now, Anthropic’s own team is publishing what this looks like in practice — agent teams building real systems inside shared codebases with Claude Code. According to YC managing partner Jared Friedman, about a quarter of YC’s Winter 2025 batch had codebases that were roughly 95% AI-generated. The gap between “AI-built” and “production-grade” has closed faster than anyone predicted.

In my world, the objection goes deeper than “AI isn’t good enough yet.” It’s “I’m not technical.” It’s “We need a real developer.” It’s “This isn’t for people like us.” If you’ve heard that last one before, it’s because it’s the same voice that’s been gatekeeping every corridor of power for generations. Who gets to build, who gets funded, who gets to walk into the room and be taken seriously — it’s the same fight wearing new clothes.

That belief is the most expensive thing you will ever hold.

Not because AI is coming for your job. Because AI is handing you the power to build solutions for the problems you’ve spent your career understanding, and you’re looking at those tools and saying “not for me.”

What I haven’t been saying loudly enough

I’ve been careful in my spaces. Measured. Polite. Matching the tenor of the room.

The thing keeping your communities stuck, the grant prospecting that takes months, the reporting that consumes 40% of your staff’s time, the program evaluation you can’t afford, the technology platform you need but can’t build, AI can address much of that. Not in five years. Now. Today.

And the thing you’ve been told you lack, the technical skill, the engineering talent, that stopped being the bottleneck. The bottleneck is now something you have in abundance: deep knowledge of the problems worth solving. The instinct to know what a good solution actually looks like, earned over years of proximity to the people you serve.

The social worker. The teacher. The community organizer. The nonprofit founder with $33,000 in the bank. You were always the ones who understood the problems best. Now you can build the solutions too.

That’s what’s actually big about this moment. Not that AI replaces workers. That AI hands the tools to people who were never given permission to build.

The urgency is real

But here’s where I agree with Shumer completely.

Right now, the future of AI is being decided by the few hundred researchers Shumer describes, at a handful of companies, building tools that reflect their world. Every month that passes without domain experts in the room is a month the architecture gets built without their values, their knowledge, their priorities.

I watched this from inside Google for nearly a decade. The hardest lesson I learned is that values retrofitted after systems are built rarely stick, and even when they are established, they’re often washed away by the pressures of the business. The same pattern I’d seen in criminal justice, in education, in economic development: by the time the people closest to the problem get a seat at the table, the architecture is already built.

Right now there is a window. The tools are accessible. The cost of building has collapsed. A single person with deep knowledge and AI tools can create in weeks what used to require a team of eight, nine months, and hundreds of thousands of dollars. That’s not a prediction. That’s my actual life.

If the people closest to the hardest problems step through this window and build, we get a different future. One where AI serves communities and not just shareholders. Where the tools reflect the values of the people who use them.

If they don’t, the window closes. And someone else builds the future for them. Again.

Think about Park Boulevard. Families on their knees in the street. Fifteen to twenty-six million people stepped into the streets that summer. Not because the facts of racial injustice were new. Because a moment arrived when stepping in felt necessary. When standing on the sideline stopped being a defensible position.

This is that kind of moment for building.

But I won’t lie to you about what it takes.

The moat called capital

The technical wall fell. But there’s still a moat, and it’s called capital.

Kindora cost about $70,000 to build and launch. Outdoorithm was around $100,000. Those numbers are a fraction of what it used to cost to start a technology company. Two years ago, this work would have required a team of engineers and half a million dollars or more. AI collapsed the cost by 80 or 90 percent.

But $70,000 is still $70,000. And if you’re a Black founder, a Latina founder, a first-generation founder whose parents didn’t own property, that money doesn’t materialize from a “friends and family round.” Friends and family rounds work when your friends and family have wealth. When they don’t, you’re stuck before you start. The racial wealth gap doesn’t pause because the cost of building software went down.

We floated credit card balances to keep Outdoorithm Collective running. Sally and I put $50,000 of personal money into equipment and operations before we paid ourselves a cent. We could do that because we had savings from years of professional-class salaries. Most of the founders I know who are closest to the hardest problems don’t have that cushion.

So when I say the wall fell, I mean it. When I say anyone can build, I believe it. But “anyone can build” and “anyone can sustain what they build” are two different statements. The second one requires capital. And right now, capital is still flowing to the same places it always has.

We’ve seen this before. After Floyd, corporations pledged $50 billion toward racial equity. Foundations opened emergency funds. A lot of it evaporated or flowed to the same institutions that always get funded. Some capital managed to reach the builders closest to the problems, but the latest backlash largely erased what was gained. Now there’s a new opening — the cost of building has collapsed, and domain experts can finally create their own solutions. But it will end the same way if capital doesn’t follow.

This is where funders, foundations, and high-capacity individuals have to step up. Not with another report on AI’s potential. With money. Foundations sitting on endowments should be opening them not just for impact investments but for venture investments in the builders closest to the problems. Individuals with means should be pooling capital into funds that back $1-5 million sustainable ventures serving real communities, not just billion-dollar unicorn bets. The incredible things our communities are capable of building won’t survive without oxygen. Capital is that oxygen.

The cost of building just dropped by an order of magnitude. If we can’t figure out how to get $50,000 to $100,000 into the hands of domain experts with working prototypes, we will have wasted the most democratizing technological shift of our lifetimes.

What I’d do this week

Shumer tells people to pay $20 a month for AI and spend an hour a day experimenting. Good advice. Here’s mine, for the communities his essay didn’t mention.

If you lead a nonprofit: Open Claude or ChatGPT this week. Not to explore. To solve one specific problem that eats your time. Feed it your last grant application and ask it to find the weaknesses. Give it your funder list and ask it to rank by fit. Start with the thing that overwhelms you most and see what happens. Prototype fast — but if you work with sensitive data, scrub it or use approved tools the same way you would with any other vendor.

If you’re a funder: Stop treating technology as overhead. Every nonprofit you fund without funding their ability to use these tools is an organization you’re funding to fall behind. Put AI capacity in the next grant cycle. Not as a special initiative. As a standard expectation. And fund the builders creating AI for communities, not just the companies building AI for profit. Consider giving startup grants or investments, not just traditional charity.

If you’ve been told you’re “not technical”: Let go of that identity. I had a C++ class in 2002 and hadn’t written real code since. Last week I built a voice AI assistant in four prompts. You don’t need to become a developer. You need to become a builder. Someone who knows what to create and can direct AI to create it. Those are different skills. You already have the hard one.

If you have deep expertise in any problem: That expertise just became the most valuable asset in the room. Not because of what you know about AI. Because of what you know about your domain that AI doesn’t. The school district. The housing system. The neighborhood. The families. That knowledge, combined with AI building tools, makes you more powerful than a team of developers who’ve never met the people you serve. Stop waiting for someone technical to build your solution. Start describing it to Claude and see what comes back.

If you’re raising kids: Shumer says rethink what you’re telling your children. My daughters are 4, 12, 14, and 16. Here’s what I’m telling them: Don’t just learn how to use AI. Learn how to develop judgment about what’s worth building and who it should serve. Study problems that move you. Get close to communities. The technical skills to build things will keep getting cheaper. The wisdom to know which things are worth building will only get more valuable. Become leaders of character who know how to connect with people deeply.

Sometimes I hold back because I wonder if maybe I’m different. Maybe my fancy degrees or Google experience give me an unfair advantage that makes this work for me but wouldn’t for others.

Then I build a voice AI assistant in four prompts and I think: there’s no way I’m the only person who can do this.

I’m not. And neither are you.

On Park Boulevard that day in 2020, nobody was polite. Nobody was measured. Nobody was matching the tenor of the room. People were on their knees in the street with their children because a moment had arrived that demanded something more than caution.

The wall is down. I’m done being polite about this. The people I care about deserve to hear it at full volume.

What’s the problem in your community you’d build a tool for — if you believed you were allowed to?

Share this with the smartest person you know who’s been told they’re “not technical enough.” They might be the most important builder of the next decade. And they need to hear that the wall just came down, before someone else builds the future without them.

Originally published on LinkedIn, 2025-12-19.

I spend most of my days building AI tools. Three startups. Coding until one or two in the morning. Shipping an entire marketplace platform over a weekend. The pace feels like a nightclub at peak hour. Relentless. Electric. Exhausting.

Then I step into social impact spaces. And it’s like walking out of a rave into a library. Everyone’s quiet. Cautious. Moving carefully between the stacks.

Same cities. Same zip codes. A world apart.

That gap isn’t just culture. It’s power.

The Funder in the Back of the Room

Last month I moderated a panel for Northern California Grantmakers alongside Beth Kanter John Kenyon and Sophie Owens. Room full of funders. Toward the end of the Q&A, a woman in the back raised her hand.

“So you’re telling me these AI models are destroying the environment. Built on extraction of copyrighted content. They infringe on privacy. They advance racist bias. What exactly is good about these? Why should I, as a funder, support anything built on these models?”

The room got quiet. She wasn’t wrong. She was naming what everyone was thinking.

The Fourth Floor of the Library

Two weeks later I was at the San Francisco Public Library, co-leading a hands-on workshop with about 30 nonprofit practitioners alongside Darian Rodriguez Heyman and Adamaka Ajaelo. The session was a partnership with Jerry Trotter from the San Francisco Office of Economic and Workforce Development and Reymon (Rey) LaChaux at the Mayor’s Office of Housing and Community Development.

Fourth floor. Wood-paneled room. Beautiful bookshelves. We had a laptop connected to a large screen, doing live prompts as people asked questions.

Then we hit privacy.

For fifteen minutes, fear filled that room. One woman was worried that information in a 990 tax form about serving immigrant communities could be used against the people they protect. Another woman works with people re-entering society after incarceration. She worried that putting any information about them into AI could violate privacy and increase their vulnerability.

I watched people’s faces as we walked through how to turn off the settings that let companies train models on your content. How to use APIs and enterprise accounts. Which companies to trust, and why trust is complicated when we’ve all seen companies violate privacy in ways that surprised us.

These concerns are legitimate. Nonprofits serve vulnerable populations. Many have fewer than 10 staff. They don’t have technical experts to vet every tool. The stakes of getting it wrong are real.

And I need to be honest: turning off training settings doesn’t mean no logging or retention. Enterprise accounts don’t mean zero risk. Assume anything you paste could be exposed someday, even if unlikely. There’s no perfectly clean option here.

And yet.

The Two Leaders in the Front Row

In the same workshop, two Latino leaders sat in the front row. They were trying to figure out how to raise funding for their communities during one of the most difficult fundraising years in memory. DEI headwinds everywhere. Funders pulling back. The political climate hostile.

We ran a live prompt together: “Help me generate a list of funders who, despite the DEI headwinds and this political climate, are still stepping up to serve the Latinx community.”

In about 30 seconds, the screen filled with 20+ organizations. Latino Community Foundation. Hispanics in Philanthropy. Hispanic Federation. Ford Foundation. MacKenzie Scott’s Yield Giving. The California Endowment. Silicon Valley Community Foundation’s LatinXCEL Fund. Each with a short rationale, which we then verified.

The two leaders were floored. What would have taken weeks of research—if they even knew where to start—had just happened in front of them.

They left encouraged. Not because AI solved their problem. But because they saw a tool that could help them navigate impossible headwinds with new ideas.

That’s the transformation I keep watching happen. Not “AI will save us.” But “oh, I can use this for X, and I know not to use it for Y.”

The Fundamental Infrastructure

AI is becoming general-purpose infrastructure. More like electricity than an app.

To generate electricity, you have to consume natural resources. You can burn coal. Burn oil. Dam rivers. Generate solar. There are trade-offs with every source. And yet we all need electricity to be productive, to participate in the economy.

You can advocate for cleaner sources. Push for better regulations. Choose providers more carefully. But refusing to flip the switch isn’t an option for most people who want to function in the modern economy. You can opt out individually. But you can’t opt out of the systems adopting it around you.

I’ve learned from two decades in corporate social impact that technology amplifies existing systems. It makes good systems better and broken systems worse. That’s why the library is right to be cautious. Broken systems—biased hiring, predatory lending, surveillance of vulnerable communities—get worse when you add AI.

But it’s also why we can’t sit this out. If we’re not shaping the applications, not demonstrating what AI can do for communities, not earning credibility through use, we lose our seat at the table when it’s time to advocate for guardrails.

It’s hard to design guardrails for a tool you’ve never used.

The Bridger’s Confession

I need to be honest about something.

When I led philanthropy at Google.org, one of my primary functions was softening the regulatory environment. Engaging policymakers. Getting elected officials to attend announcements. Providing quotes for press releases that signaled these companies have society’s best interests at heart.

I participated in that machine. I strove to hold the tension between company benefit and community benefit. I didn’t always do it perfectly. Some projects I worked on provided a strong halo for the company and reduced regulatory pressure in ways that benefited Google more than the communities we claimed to serve.

That’s part of being a bridger. If you’re going to bridge, you’re going to get your hands dirty. You can’t be ideologically pure. You can strive to balance those things well. You won’t always succeed.

I say this because I think we need more bridgers. People willing to live in that tension. People who refuse to give into tech utopia but who also aren’t afraid to engage. Folks who understand the externalities but won’t sit it out while others reshape the economy.

If we allow the private sector to completely monopolize this tool, we’re worse off. We need to use it. We need to earn our seat at the table. That means getting our hands dirty.

The Risk Ladder

Here’s the practical framework I’ve been sharing in these workshops. Think of it as a ladder:

Level 1 — Low risk (public-facing content): Draft newsletters. Rewrite emails. Summarize public reports. Generate social media posts. Research funders. None of this involves sensitive data.

Level 2 — Internal, low sensitivity: Meeting recordings and summaries with team consent. Job descriptions. Strategy document updates. Internal communications. Still no client data.

Level 3 — Sensitive, only with guardrails: Enterprise tools only. Data minimization. Redact identifying information before pasting. Documented policies about what goes in and what doesn’t.

Never: Client-identifying information or legally sensitive data in consumer chat tools. Period.

The underlying mindset is simple: walk through your day asking, “This task takes me a lot of time. Can AI help?” Then check which level it falls on.

Most of what overwhelms nonprofit staff—grant research, meeting notes, donor communications, strategy docs that are months out of date—lives in Levels 1 and 2. That’s where you start.

The Ask

We need people with social conscience and proximity to social problems engaging these tools. Driving solutions. Shaping the public conversation. Finding applications the private sector won’t prioritize.

The library is where we ask the questions the nightclub doesn’t have time for. We need that caution. But if all we’re doing is sitting in the stacks while everyone else is in the dance, that’s not where we need to be right now.

Our communities are moving slower than the moment demands. That pace is a rational response to high stakes and low capacity. But it’s also leaving us behind while the private sector rewrites the rules of the economy.

Every time I step from the nightclub into the library, the contrast gets more jarring. Not because the concerns in the library are illegitimate. They’re not. But because the music in the nightclub isn’t stopping.

So here’s my question: What’s one thing you could try in the next 30 days? One task that takes too long, one workflow that’s overwhelming, one place where you’d be willing to flip the switch and see what happens?

I’d love to hear what you’re experimenting with, or what’s still holding you back.

Originally published on LinkedIn, 2025-10-30.

OpenAI announced this week that their nonprofit now controls a $130 billion stake of the company, one of the largest philanthropic fortunes in America.

I spent a decade giving away hundreds of millions in corporate philanthropy. So I know what comes next.

Carnegie built libraries after crushing workers. Rockefeller funded Black colleges after extracting oil wealth.

The libraries still stand. So does the inequality.

Here’s what makes OpenAI different. They didn’t extract oil or steel. They mined something more diffuse — every book ever written, every photo ever shared, every line of code ever posted — and turned it into private property.

As OpenAI establishes its foundation, the question I can’t quite answer: Does the money us philanthropists help move to communities outweigh the permission we give the system to keep taking?

Most days, I don’t know. But here’s what I do know.

The Window Is Closing

For decades, Silicon Valley had one primary advantage: they knew how to code. You didn’t.

That’s over.

What used to take a team of eight engineers nine months and hundreds of thousands of dollars now takes one person with AI a few weeks. It’s a special kind of magic.

As AI makes coding routine, what becomes valuable is judgment—knowing what’s worth building because you’ve lived the problem.

But only if we move money and decision-making power to people close to real problems before all the wealth piles up in the same hands again.

What if capital, computing power, and control flowed to people with lived experience instead of computer science degrees?

Three OpenAI Moves That Would Actually Mean Something

1. Empower People Who Know Problems To Build Solutions

90-day programs where people close to real problems learn to direct AI to build tools. Not coding bootcamps. Not startup accelerators.

The social worker who’s watched families fall through intake cracks for 15 years? Give her free access to AI, a session with OpenAI’s team, and a safety review before launch.

2. Give Money That Comes Back

But don’t just stop at product. Nothing gets built without capital. $25K-$100K grants to 20,000 ventures over the next two years. If they become sustainable, they pay it back and it funds the next group. If they don’t, it converts to a gift.

I built Kindora—an AI fundraising platform for nonprofits—using AI tools like Cursor and Claude Code. Total cost to build and launch as a public benefit corporation: $60K. It’s already serving hundreds of organizations that can’t afford the $4,000/year platforms that give them thousands of overwhelming matches.

Traditional venture capital wants billion-dollar exits. Our community-led ventures might be $1-5M sustainable businesses serving users that desperately need them. That’s not failure. That’s success serving the right people.

3. Take the Risk Banks Won’t

$500M to absorb the first losses on loans. When you take the first 10% hit, community banks like CDFIs will fund the next $5-10 billion.

That’s building infrastructure that outlasts any single grant. Creating markets, not just funding companies.

Five Questions That Tell You If It’s Real

Here are five questions to determine whether OpenAI is earning the legitimacy its philanthropy will bestow upon it:

1. When profit and purpose conflict, which wins? Not whether you “considered” community impact. Whether you can point to a specific decision where you chose community wellbeing over making more money.

2. Are you funding relief or changing the systems that create the need for relief? Charity eases suffering. Justice asks why people are suffering in the first place.

3. Do you measure social impact as seriously as you measure business results? Or is impact vague and directional while you track brand metrics to three decimal places?

4. When you fund innovation, are you solving for technology or solving for power? Are you funding apps or funding the messy work of shifting who gets to decide?

5. Do affected communities actually shape your work—or are you designing solutions for them? Getting close to people requires relationships that change you, not just programs you control.

The Choice Before Us

Taking OpenAI’s money to run their philanthropy would mean your name, your work, your thinking becoming part of how they tell their story. Anyone who takes this job becomes the permission slip they’re seeking.

I’m not available. When my wife and I were running Outdoorithm Collective on credit card float, I built Kindora because we needed it. Somewhere in that process, I stopped waiting for institutions to do better.

But here’s the question I’m not asking OpenAI, and probably should be: Are you willing to forgo the activities that require redemptive philanthropy? Not “how generously will you give back?” but “what will you refuse to build because it can’t be made right through subsequent charity?”

Until we’re asking that question, we’re just negotiating the terms of extraction.

And here’s the question I cannot ask, because my professional identity depends on not asking it: Should the role of “thoughtful philanthropy leader” exist at all?

Originally published on LinkedIn, 2025-10-16.

The day after the Kehinde Wiley exhibition opened at the de Young Museum, my phone lit up. A Google policy lead needed to talk—now. A senior San Francisco official had pulled her aside for a take down: one of our Oakland organizers had criticized the administration in a recent speech. We’d brought that group into the program. Funded them at a significant level.

“Fix this,” was the subtext. “Make it go away.”

Welcome to the kitchen of philanthropy, where everyone sitting in your dining room has different tastes.

The check is easy. The change is hard.

Live in the Tension

To understand that phone call, you need to understand the night before.

March 2023. I’m at the de Young Museum with Darren Walker and Kehinde Wiley to open his An Archaeology of Silence exhibit—25 monumental works of fallen Black figures. In the main gallery, Googlers beam, proud to be part of the largest corporate foundation grant in the museum’s history. Wine flows. Executives shake hands with museum board members.

Thomas Campbell (FAMSF CEO), Kehinde Wiley (artist), Darren Walker (Ford Foundation President), and Justin Steele (author) at the opening of An Archaeology of Silence, March 2023

In an adjacent room, Oakland organizers prepare healing circles for communities who’ve lost loved ones to violence. Same building, different worlds.

This is the setup we’d carefully orchestrated with Abram Jackson, bringing both constituencies to the same table. The museum got its blockbuster exhibit and free-admission weekends. The community got grief workshops and a speaker series. Google got to stand at the intersection, claiming we could hold both.

For one glittering evening, it worked. The contradictions held. The tension felt productive, even beautiful.

Then morning came, and with it, that phone call.

The Dining Room Is Full

Philanthropy is a twelve-course tasting menu with a crowded dining room.

Your principals —board members, executives, family offices—control the resources. They want familiar flavors. A glass of cabernet and a filet.

The community facing the deepest challenges sits at another table. They need something real. The soul food gumbo that actually sustains—not always pretty or plated right, but the thing that nourishes.

Add the media table. The staff table. The political players who weren’t invited but somehow have opinions about your menu.

Your job? Cook something real that travels between tables—that works in the boardroom and the community center.

The Kehinde Wiley Course

Months before that March evening, Adrian Schurr and I walked into the de Young Museum having rehearsed a polite decline. Then we heard Wiley’s vision: a body of work confronting systemic violence against Black bodies, brought to millions through the Google Arts & Culture team, turning a museum exhibition into digital narrative change.

The museum asked for free‑admission weekends, grief workshops, a respite room. We knew the exhibition needed community voices, too. So we funded healing circles and a speaker series.

The policy team wanted the community grant removed. Keep it simple. Keep it safe.

Had we pulled the community grant, the museum would have had its carefully curated experience. Google would have had a brand moment. And the communities actually living with the violence Wiley’s art depicted? They would have remained what they’ve always been in these spaces—subjects to be discussed rather than partners to be heard.

What Good Chefs Actually Do

After a decade cooking these impossible meals, here’s what I found works:

1. Balance what you serve

Treat your portfolio like a tasting menu with three movements.

The opening courses: Light, accessible, beautifully plated. Nothing challenging. In corporate philanthropy, that’s computer classes for kids, disaster relief, veterans programs.

The bridge courses: The dishes everyone knows, transformed. In corporate philanthropy, that’s cybersecurity clinics at HBCUs, job training that quietly includes returning citizens, mental health programs in schools that happen to center immigrant families.

The brave courses: What separates brave chefs from conventional ones—serving what makes diners pause, then realize this is what they’ve been missing. In philanthropy, that’s programs that challenge accepted narratives that underpin our social hierarchy, proposals that aim to redistribute concentrated wealth.

2. Time your service

Don’t bring out the hardest dishes first. Your first 90 days? Start with the inventory. What ingredients are actually in your pantry? Who’s been burned by the last chef? Which dinner guests have the CEO’s ear, and what did they last complain about? Map the real kitchen, not the one in the brochure. Then cook one small, perfect dish that everyone can taste. Build from there.

3. Learn what nobody will tell you

No one hands you Things That Will Get You Fired. Watch for what tightens your CEO’s jaw. Notice which words make legal perk up. Stay close to the exec who hired you—they’re your thermometer; if they start sweating, you’ve turned up the heat too high. Build a mental map of no‑go zones.

4. Bring dishes to the table

A meal nobody sees might as well not exist. Don’t hide your best dishes in the kitchen. Bring grantees into the room. Document with photos, stories, and impact metrics—not for the menu, but to prove the cuisine transforms lives.

Why Most Kitchens Serve Forgettable Meals

The money exists. The intent exists. The need is urgent.

What’s missing? The bridgers. Chefs who cook across cultures. Who can take that unfamiliar ingredient from the community’s garden and make it sing on the principal’s plate.

They’re the chefs who’ve earned their stripes in both worlds: maybe they’ve run a nonprofit AND closed deals, organized communities AND read P&Ls, or lived the problems AND speak the language of solutions. They know when to say “ROI” and when to say “justice”—and understand these aren’t always different things.

Most kitchens serve safe, forgettable meals. Principals get their steak. Community gets something. Boxes checked. Little transformed.

After Service

The Wiley exhibition opened to 35,000 visitors during free weekends. The grief workshops filled. Oakland organizers and the San Francisco museum found common ground.

We notched real wins. But not everyone saw them that way.

That tension never fully resolves. Serving company interests and serving community needs don’t always align perfectly. Someone has to hold that complexity.

My team learned to cook in the tension. They saw that the best work happens precisely in that uncomfortable space between safe and right.

Your Kitchen Awaits

So you’re stepping into your first giving portfolio. The dining room is full. Orders conflict. The heat is rising.

Good.

Live in the tension. If you ever stop feeling it, you’re probably not doing the work. This isn’t about writing checks. It’s about feeding people across impossible divides—making trust travel farther than suspicion—cooking meals that transform both the fed and the feeder.

Welcome, chef. Now cook something that matters.

Originally published on LinkedIn, 2025-02-20.

After two decades navigating corporate social responsibility work and an unexpected role change, I recently stepped away from big tech philanthropy to build something new. The decision of where to journey next wasn’t easy: I’ve seen firsthand the transformative power of institutions committed to social impact. To change culture, we must create culture—and institutions remain our most potent culture-makers.

But I’ve also witnessed how institutions resist deep transformation. These systems are like houses built long before we arrived—you can often paint the walls, but you may not touch the beams.

This feels especially urgent right now. As democracy frays, as vulnerable communities are pushed aside, as artificial intelligence promises to amplify both our creative and destructive capabilities, I see many others wrestling with fundamental questions:

Do we stay and cultivate a faithful presence within these institutions? Do we imagine something new and find the courage to go build it? Or do we perhaps try the narrowest path of all and attempt to navigate both?

The Depleted Soil of Institutions

When I started in corporate social responsibility, the institutional soil wasn’t perfect, but it was fertile. Seeds of change were growing, budding with possibility.

Consider my team’s partnership with Bryan Stevenson and the Equal Justice Initiative to confront America’s history of racial terror lynching. By bringing together Google’s technical expertise and resources with EJI’s moral authority and research, we helped legitimize this difficult truth-telling in the eyes of mainstream Americans, reaching over 35 million people through our digital platform. This was precisely the kind of initiative that demonstrated how institutional power could be redirected toward genuine culture change.

But today, that soil in many of our nation’s most powerful institutions has been depleted. What was once a garden with much potential has become increasingly hostile to initiatives that might genuinely renew society or build true solidarity across communities. When institutions become purely transactional spaces, they lose their capacity to cultivate civic virtue and meaningful social bonds.

The Limits of Faithful Presence

Let me say it straight: Inside many of our most dominant economic and political institutions, faithful presence alone is no longer sufficient. While some institutions may still maintain space for genuine transformation, those working within them must take an unflinching look at whether they are changing these systems or being changed by them.

The hardest truth I’ve learned this past year is that far more change agents are being absorbed (or even ejected) by our powerful institutions than are transforming them. We tell ourselves we’re making incremental progress while core power structures remain untouched.

The tough reality is that the system does not step aside because you have made a better argument. It does not surrender because you have planted seeds in the cracks. It will often watch you plant new things in its garden, admire your patience, nod at your discipline, and then pave over what you’ve planted the moment it becomes inconvenient.

A Call to Build

Given this reality, many change-makers are asking a different question in this moment: If we cannot transform these institutions from within, what might we build instead? Many of us will find our most productive space in the underground.

For example, consider how hip-hop’s most revolutionary voices emerged from underground spaces. While many artists later engaged with mainstream culture, their initial power came from building something entirely new, unconstrained by existing systems. The underground wasn’t just an alternative—it was a protected space where different values could take root and flourish before engaging with mainstream power structures.

For those of us stepping out to build new models, I believe four foundational values are essential:

• Power : Not as domination, but as the capacity to protect and sustain transformative change—because new ways of being must be defended until they can take root and flourish.

• Love : Not as sentiment, but as transformative action—because if we do not center love, then whatever we build will crumble beneath the weight of the same hierarchies, the same violences, the same exclusions that have always existed.

• Truth : Not as a weapon, but as a foundation—because we cannot build new institutions on foundations of unacknowledged harm, nor can we weaponize truth in ways that dehumanize others.

• Solidarity : Not as slogan, but as the deep practice of shared struggle and mutual commitment across ALL people, without exception—because divided movements become conquered movements

The Leadership We Need

These values aren’t just principles—they’re the foundation for a different kind of leadership. The work of building new models demands leaders who deeply understand power—not to perpetuate it, but to transform it. People who can imagine a different conception of institutional purpose – one that places collective flourishing and moral values at the center rather than treating them as optional add-ons to market efficiency.

These leaders face a difficult path: their commitment to transformation will put them in tension with existing power structures. Their careers, relationships, and security may be put at risk. But this willingness to put one’s comfort and status on the line in service of building something better is precisely what gives their leadership its transformative power.

The Choice Before Us

Whether you choose to practice ethical navigation within institutions, join an underground movement to imagine and build something new, or take the road less traveled and attempt to navigate both, the key is clarity about our ultimate aim: creating systems that serve all of humanity, where everyone can flourish.

In a world increasingly driven by winner-take-all logic, perhaps the most urgent question is this:

What sacred line are you quietly watching erode that you could defend today? And more importantly, what are you prepared to do if you see it vanish entirely? Will you mourn? Will you resist? Or will you build something that cannot be so easily erased?

The soil may be depleted in our current institutions, but the earth beneath them remains fertile with possibility – if only we have the courage to dig deep enough to reach it.

I’d love to hear your experiences navigating these choices between institutional transformation and building alternatives. What’s working? Where are you finding hope and renewed imagination?