Our approach

Do the most good with what we have.

We are a small foundation. That makes one question unavoidable: with the money and time we have, how do we do the most good?

Our answer has two parts. We give the way effective altruism recommends: we choose by evidence, will count the cost of each result, and will publish what didn’t work. And we use AI and modern software to make each hour and each dollar go further than it could five years ago.

Effective altruism, the way we practice it

Four habits behind every decision.

Effective altruism asks how to do the most good with the resources you have. We take the question seriously, and it shapes every program we choose.

  • 01Big, neglected, solvable

    We look for problems that affect many people, that others are not already solving, and where our help can make a real difference.

  • 02Every dollar counts

    We will track the cost of each result, and we favor work whose cost for each extra person falls toward zero, like a free app or a platform. That is why two of our three programs are things we build.

  • 03Would it happen without us?

    We fund what would not otherwise happen. If someone else would pay for it anyway, our money belongs somewhere else.

  • 04We will publish what didn’t work

    Every program will report what it delivered, what changed and what lasted, including the disappointments. How we measure

Loyalty and memory. Some of our giving, like our support for the Armenian community, is also an act of loyalty to Rouben and his heritage. We say so openly, and we hold that work to the same standard of evidence as everything else.

How we choose

Four tests before we build.

An idea becomes a Foundation program only when it passes all four.

  1. 01

    It closes a real gap

    One that stops talent from becoming opportunity, and that no one else is closing well.

  2. 02

    We can see it fully paid for

    So it serves people on the strength of their need, not their ability to pay.

  3. 03

    It is built to last

    Not whether it can launch, but whether it will still be serving people in twenty years.

  4. 04

    It can be held to account

    It reports what it delivered, what changed and what lasted.

See the programs that passed

How we run every program

Five commitments.

  • 01We pay for what we start

    We would rather back a few programs completely than many in part. When we take one on, we make sure it is fully paid for, from our own funds and from donors who pledge to it, so it can put its mission ahead of revenue.

  • 02We measure honestly

    What we delivered, what changed and what lasted, reported alongside what did not work.

  • 03We stay close

    Our community programs are run by people who know the families they serve, and our apps and platforms put a person in charge of every judgment call.

  • 04We protect independence

    Where we support communities we do not govern, such as the Menzies Alumni Group, we strengthen what they build without touching their decisions.

  • 05We never sell personal data

    We never use the details of the people we help for fundraising, and we keep only what we need to help them.

How we are governed

AI as a force for good

A small team, with far longer reach.

Powerful AI could help the world make decades of progress in health, poverty and opportunity in a fraction of the time, but only if people deliberately put it to work for those goals.

That view is set out by Dario Amodei, Anthropic’s chief executive, in his 2024 essay Machines of Loving Grace, and we share it. Anthropic has put resources behind it, with discounted access to Claude through Claude for Nonprofits and, in June 2026, Claude Corps, an initial $150 million to train 1,000 fellows and match them with nonprofits across America.

We believe the same holds at the scale of a community. AI already lets a foundation our size do things that were out of reach five years ago.

  • Admit PathOne expert, many families

    Statistical models compare each applicant with students actually admitted, while trained readers make the judgment calls. Guidance that private consultants charge far more for starts at $0.

  • App StudioBuilding what was never viable

    AI tools have cut the cost of making a polished game for blind players to a level a foundation can afford.

  • Every programMore time with people

    When AI handles routine paperwork, our small team spends its hours with families, students and partners.

  • TrainingNew skills for a changing economy

    The same tools that are changing work can retrain the people affected. App Studio is planning a training program to do that; its first cohort has not started yet.

See how AI works in each program

How we use AI responsibly

Careful, and in the open.

The companies building AI, Anthropic among them, are clear that it carries real risks. We take that seriously.

  • 01People make the judgment calls

    On Admit Path, sensitive topics are escalated to a human reader, and independent trained readers score essays, recommendations and mock interviews.

  • 02Privacy comes first

    [TBC: our rule on putting personal details into AI tools, as actually practiced]

  • 03We are honest about limits

    We tell people when they are using an AI tool, and we never present an AI’s guess as fact.

  • 04We share what we learn

    So other community organizations can use AI well too. If you run one and want to compare notes, write to us.

We build with Claude, made by Anthropic. Nothing here suggests that Anthropic endorses the Foundation or its views.

Ask us how we use AI

Hold us to it

Back work that goes further.

Give to a program, sponsor a free app, or ask us hard questions about how we measure. We will answer them.