Introducing Career Workshops for AI Safety Grantmaking

If the field needs more grantmakers so badly, where is a junior person supposed to fit? It is easy to read all of this and come away with the sense that, yes, this is important, but also obviously a job for someone older, more experienced, and more “strategically sophisticated” than you might be.

If you are just entering the field, the honest answer is that you probably do not yet have the strategic taste to set a grantmaking team's highest-level priorities. That is normal. Most people do not. But that is a very different claim from saying you cannot be useful. There is a big gap between “should not decide the whole portfolio” and “has nothing to contribute.”

The field is especially labor-constrained in the active parts of grantmaking, not just in reading proposals once they arrive. The part of grantmaking most people picture—reading a proposal, deciding yes or no—is also the part most easily delegated. What cannot be delegated as easily is the active work: figuring out what should exist, mapping who is already doing related work, pressure-testing a theory of change, scoping a budget, and helping turn a half-formed idea into a fundable project.

Jake Mendel at CG has said he spends the majority of his time not evaluating proposals but trying to get better ones into the pipeline. This is the work that is most acutely bottlenecked.

Traditional high-stakes, professional field has solved this problem: medicine has residencies; law has clerkships; investment banking has analyst programs. Equally as much as being places that build experience, these are infrastructure that makes the senior professionals more productive. Instead of operating on a patient, the first-year surgical resident assists in the procedure—enabling the surgeon to operate on more patients.

Grantmaking has no equivalent: there is no structured way for a junior person to enter this field, contribute meaningfully, and develop judgment over time. And we've borne the consequence for this gap: while senior grantmakers remain overloaded, promising junior people might never even discover this path. The consequence for AI Safety is then that good opportunities are left on the table, and worst of all, not even because of the money.

A junior person does not need to set the priorities. They need to help a senior person act on theirs faster. That is a very different bar, and a much more reachable one. Your job does not have to be coming up with that direction from scratch. It can be going out into the world and making it more concrete: talking to orgs, gathering signals, drafting the memo, spotting overlap, finding the right people, surfacing blockers, and helping move an idea from “someone should look into this” to something real.

And the nice thing about starting there is that the learning is built into the work. As you do it, your taste gets better. You start to see which orgs are stronger than they look, which people actually follow through, where the field is crowded, and where it is weirdly empty. You also build relationships across the ecosystem almost by accident. Even if you later leave grantmaking, you keep the map.

That is part of why we think of this as a lever, not just a job. Good judgment here moves more than your own output. It changes what gets funded, what gets built, and which problems receive real attention. But the path into that leverage is much less mysterious than it first appears. You do not start by controlling millions of dollars. You start by becoming useful to someone who is already trying to direct them well.

What To Do Next?

So what does that actually look like in practice? The lowest-cost experiment I know is to do small pieces of work that look like the actual job before anyone gives you permission. One version is to write up a more formal version of your own reverse-engineering and send it to current grantmakers: what did I miss, what would you rank as most useful, where do different teams actually spend their time? Another is to map the current grantmaking orgs, what they each do, where their comparative advantages differ, and which gaps seem most real.

More broadly, the design principle is simple: practice on the actual units of work the job consists of, just without the money. Write a recommendation memo on a real public proposal. Do a landscape memo on university groups and ask what is missing. Write an Epoch or AI 2027 synthesis that ends with, so what should change in current grantmaking? Design a lean $300K fellowship budget. Draft an RFP in an underfunded area. These are small enough to be doable and real enough to teach you something.

The point of these projects is not just that they help you learn. They also produce artifacts. That matters. A finished memo, budget, or RFP is something a current grantmaker can react to. It gives them a concrete object to critique instead of a vague expression of interest. It also gives you something you can own, revise, and point to later when you want to have a serious conversation.

They are better done with other people, too. A small group can split interviews for a landscape memo, compare judgments on a proposal, or stress-test each other's budget assumptions. That gives you accountability, lets you meet other people interested in this path, and makes the whole thing feel less like private career flailing and more like the beginning of a real practice.

From there, a few entry points open up. You can join a larger grantmaking or donor-advisory org and learn from the inside. You can become useful to independent donors who want a giving strategy that does not simply duplicate Coefficient or Longview. You can start smaller by regranting on Manifund, helping run a tactical fund, or supporting someone else in scoping and seeding a project. There is no single ladder here. The common denominator is that you become legible by producing good judgment in public.

There are also some immediate actions that matter right now. If you know someone at Anthropic who is thinking seriously about giving, connecting them with existing advisors before the IPO is probably much more useful than waiting until after liquidity. If you are already a donor, funding grantmaking capacity itself may be one of the highest-leverage bets available. The field does not only need more capital. It needs more people who can help direct it.