Lately, I’ve been thinking a lot about the design of grant programs, from small microgrants to regranting to ambitious new platforms to galaxy-brained schemes for impact markets. Here are five components that I think any grant program needs to be good:
1. Money
This one is obvious: a grant program needs money to give out.
Historically in EA and AI safety, this started as individual small donations from earning-to-give, to Dustin Moskovitz’s money via Good Ventures, to an ill-fated boom around the FTX Future Fund, and now everyone preparing for the frothy Anthropic and OpenAI dollars.
2. Taste
This is also kind of obvious. Most people’s image of “what makes a good grantmaker” is “excellent taste”, which is to say, the ability to discern between good or bad projects.
Grant taste comes in a few forms:
- Taste in people: founders of projects, leaders of orgs
- Taste in ideas: whether a particular idea might succeed; whether it’d be good if it did
- Taste in fields: cause prio across technical AI safety vs policy vs fieldbuilding
One problem with taste is that everyone invariably thinks that they have good taste. Also, there isn’t necessarily One True Taste, so it can be a bit confusing to think about “better” or “worse” taste. Money might instead consider whether one particular Taste is aligned with her values.
How do you improve your taste? Probably: doing similar work yourself; seeing many examples; getting feedback from peers or a mentor; watching grant results over time.
3. Dealflow
“Dealflow” is a term of art among startup investors, to refer to the funding opportunities (”deals”) that come across your table.
Good dealflow can look like:
- Seeing projects that other funders wouldn’t
- Seeing many more projects than other funders
- Good opportunities contact you first, or preferentially accept your funding
- Creating opportunities from scratch
“Improving dealflow” currently feels underrated among philanthropic funders. Compare the shenanigans VCs pull to generate dealflow (sponsor events, host podcasts, post thinkpieces, ask for intros), versus what EA funders do (…put up an RFP?).
Ways to improve dealflow include: reach outside your existing scene; invest into comms and marketing; offer a good grantee experience for retention + referrals; provide value beyond just Money (such as connections, advice, intros); share dealflow between other funders; chase hard on great founders.
4. Hustle
“Hustle” is my catch-all term for the legwork that goes into coordinating between money, taste, and dealflow; it also covers the logistics that go into getting the grants to the grantees. People with money or taste are often incredibly time-poor; hustle is the glue that keeps the grant program together.
The word hustle may conjure an image of some shady car salesman hustling to sell you a piece of junk. But hustle doesn’t have to look like that. It’s often a polite follow up, being relentless about moving the ball forward, in a way that your counterparties would endorse.
It doesn’t even have to come from a human; you could automate hustle with infrastructure that reduces the number of human touchpoints required, as we do on Manifund via digital grant agreements and self-serve payouts over Stripe Connect.
I think hustle is also currently quite underrated. Parts of EA & AI safety feel like a competition to demonstrate the highest IQ, or write the best-argued blog post — a culture inherited from academic norms. There’s a lot of room for generalists to contribute, by coordinating well and moving money quickly.
5. Trust
Trust, like hustle, is more of a meta component. If the different components above don’t trust each other, no grants will be made.
Ben Kuhn identifies trust as a bottleneck to growing teams quickly; across AI safety philanthropy, I expect trust will likewise be the bottleneck to deploy this wave of funding. The AI safety community has started with a lot of mutual trust due to being small, weird, demanding, and socially interwoven; this probably won’t scale.
Oli Habryka has spoken about the “cursed game of philanthropy”: the lack of trust behind the principal (Money) and agent (Taste) in any kind of philanthropic system. They’re designing Lightcone Commons to require less trust, by placing final funding decisions in the hands of money; I’m interested in seeing how that plays out.
Another way to resolve the trust bottleneck is to colocate Money and Taste in a single person. Leo Gao’s microgrants, giving out his own money from OpenAI, is one currently operating example. I’m excited for the Anthropic and OpenAI windfalls to empower more employees, and hope that they don’t just delegate it all away to funds. I like AI safety regranting because it likewise colocates Money with Taste or Dealflow, to individuals with a strong track record in AI safety nonprofits.
I’m also interested in other ways of increasing trust across the funding ecosystem. Manifund’s commitment to transparency is one way we push for more trust; we’d love to see other funders ask applicants to publish their project proposals, and in turn publish how much money they grant. Other ways to foster trust include in-person events and hubs; social vouching; writeups, retros, reasoning transparency, shared mental models; establishing track records of taste.
Breaking down some recent grant programs
ACX Grants
- Money: Mostly from well-monied patrons who are ACX readers; about a quarter from Scott directly.
- Taste: Scott Alexander, now increasingly via judges he recruits to help review proposals
- Dealflow: ACX readers
- Hustle: Scott sending a zillion emails and generally coordinating everything himself. (More recently, Manifund helping with payout logistics)
- Trust: Banked up in Scott over twenty years of blogging
- Money: Anton Makiiesvkyi put up $1m of his own money from earning to give
- Taste: Delegated to Manifund regrantors Gavin Leech, Marcus Abramovitch, and Ryan Kidd
- Dealflow: launched on LessWrong/EA Forum/Manifund/AISafety.com
- Hustle: Matt Brooks built the site, Melissa Samworth designed it, lots of legwork by Anton
- Trust: The team having collaborated for a while
Lightcone Commons
- Money: Jaan, Dustin, a mix of smaller donors and funds
- Taste: Evaluators like Zvi, Eliezer, Nate Soares et al
- Dealflow: a launch announcement, maybe more marketing to come?
- Hustle: Oli & the Lightcone team
- Trust: Banked up in Oli, having succeeded at various projects like LessWrong and Lighthaven, and more topically Lightspeed Grants and SFF
Coefficient Giving
- Money: Mostly via Dustin/Good Ventures, but now soliciting external donors
- Taste: a small army of specialized grantmakers across different sub-funds
- Dealflow: mix of RFPs and active grantmaking. Because they have a lot of money, basically all big orgs seek them out.
- Hustle: another army of ops/logistics people
- Trust: Originally inherited via Givewell, now stands as an institution in itself
Can money buy taste/dealflow/hustle/trust?
In some ways, yes:
- You can offer grantmakers a high salary and get better talent. (But beware principal-agent problems.)
- If you’re known to have a lot of money, the deals will come to you (for better or for worse). You can also buy dealflow via paid referrals and advertising.
- You can likewise hire people with hustle, or pay to develop systems that automate grant logistics
- You can design incentive systems that encourage trust between actors, eg Taste a % cut of Money granted. (Yes, this might incentivize recommending more $; no system is perfect.)
But also, it’s sometimes hard or expensive or suboptimal to directly convert money into things you want. There’s a lot of alpha in finding shortcuts.
So: if you have money, and want to give it away: think about what components you have and need. If there’s someone you trust with taste, dealflow, and hustle, you could simply hire them to do it.
If you have moderate trust in someone else’s taste and want to outsource the dealflow and hustle, come talk to us — this is what Manifund specializes in!
Appendix
Title inspired by Holden Karnofsky’s Rowing, Steering, Anchoring, Equity, Mutiny
I also love these reflections from new grantmakers:
Nice framework!
I have one important caveat for the topic of dealflow. While I agree that the EA community has a lot to learn from VCs, there's one crucial difference: startup funding has a much larger competitive component.
A for-profit investor wants the company they invest in to succeed, while maintaining an equity stake in it to profit from that success. This means they can win only if the company accepts their investment.
In contrast, a non-profit grantmaker is (or at least should be) concerned almost exclusively with helping the organization succeed. Whether it's their or someone else's money isn't very relevant.
This makes some lessons to be learned from the VC community irrelevant or actively harmful (if grantmakers start competing for the same projects). At the same time, it enables a level of collaboration and common infrastructure that wouldn't make sense in the for-profit world and which needs its playbook written from scratch.
Thanks for the thoughts! I think we might disagree somewhat strongly - I actually think that nonprofit grantmakers should approximate their own assessment of impact as a function of "what fraction of the best nonprofits did I fund? For what size? How early?" - aka the same considerations that a for-profit vc considers for return on investment.
I agree that, at the end of the day, we all care about total impact in the world, but I claim that "whose money funded the thing" is basically the most important thing for us to consider when we're trying to decide which grantmakers are good. (obviously other kinds of support can matter too, such as advice and connections, but those are not as central)
As an analogy, if we're allocating impact credit for a project between humans, "whose time was spent on the thing", "whose expertise was most needed for the thing" are natural and good examples of considerations we'd use to decide who was having the impact.
Why does it matter "who was having the impact"? Because without working through such problems, it's really easy to get into double or triple counting problems, where both grantee and grantmaker (and eg grantmaker's funders like Good Ventures for CG) are all implicitly taking credit for the impact. This leads to misallocation of our scarce talent and money.
As a bit of a tangent:
This approach only makes sense if you consider your available funding as a VC/grantmaker to be ~unlimited. In the VC world this might work for a16z or YC. Your regular VC fund/angel investor doesn't ultimately care what fraction of good deals they pass on or how early they joined on a deal but what literal ROI they get: net profit divided by money invested.
Sorry, yeah, mostly I was using fancy language to describe ROI; or, how a vc might try to improve/optimize their ROI: work hard to get access to good opportunities at an earlier stage.
(which is exactly what I'd like to see grantmakers doing more of)
I fully agree that attributing the impact is very important as a signaling mechanism. However I don't get why we should prioritize "who gave the money" over "who identified the opportunity" for this attribution.
Admittedly, I have zero background in non-profit grantmaking. I'm coming from the for-profit world of software products. In my world what counts as impact is counterfactual outcomes, not resources spent.
Imagine that John, the leader of team A, identifies a gap that is turning customers away, gets team B to deliver a fix and as a result the sales double. I'd say that John made the biggest counterfactual impact here, even if he didn't lift a finger on the delivery part. Without further context, I'd also consider John leveraging team B, rather than his own, as a positive signal, showcasing that he can think outside the box to deliver desirable outcomes.
Yeah, such solution does lead to more credit being claimed in total for the same work across the entire organization. It's the inherent cost of collaboration on anything non-trivial. There's very few environments where that cost would outweigh the benefits of collaboration and, at least in my intuition, non-profit grantmaking ecosystem isn't likely to be one of them.
Sure! Like to be clear, there's a very interesting meta, strategic question of "how should we conceive of the EA ecosystem? Are we all just on the same team? Should we act like we're all colleagues of a large for profit org?"
And, it wouldn't be crazy. I think to date, aspects of EA and AI safety act like this. And there's something beautiful, admirable about the attitude of the "don't worry about whose name is first on the paper, or whose getting paid the most, so long as the job gets done".
But otoh, consider the Coasean question, why do firms exist? Specifically, why are there multiple different firms? The market is efficient because the firms are competing for dollars; in our economy/society, dollars are how we divide up "credit", how we represent a claim on current or future work by others.
My general thesis is: EA has been historically small and collaborative, almost Dunbar sized. But AI is scaling up fast, and so AI safety needs to scale up; and also, AI money means that the rest of EA can also scale up. And we've seen repeatedly that markets are the best way to scale complex collaboration, compared to eg top down central planning. (Thus, we should implement impact markets.)
For more on this thesis, see the rest of our newsletter! https://manifund.substack.com/
I think we're talking past each other on one specific thing: input dollars and output dollars aren't the same dollars.
Firms compete for earned dollars. That's the output, and it's what credits the firm for the value it creates (by spending money, customers signal that they value what the firm provides above its monetary cost). Invested dollars are an input. They happen to share a unit with the output, but that doesn't mean that they measure the same thing.
I read the recent substack post on impact markets and I think this is where it runs into trouble. An org's valuation is implied by how much money it raises later. This measures how well a funder anticipated other funders, not the actual impact.
At a surface level this appears congruent with how the for-profit world operates. Startup valuations are also set by the money raised in later rounds. However there's a crucial difference - the value of firms is ultimately grounded in the expectations of future output in a way that value of non-profits can't be. Equity prices are guesses about very concrete future cash flows from people buying things the firm offers. The accuracy of those guesses can be ultimately checked against something that isn't another investor's opinion. The output is uniform - a dollar earned on a grocery store chain is the same as a dollar earned on an AI company[1].
In contrast, EA orgs aim for "positive impact" which can be quantified only by a very subjective evaluation with multiple assumptions, not by any market mechanism. The output is not uniform. Even if we had perfect measurement (we don't) I might value 1 human QALY above 1000 shrimp QALYs while you do the opposite.
This results in money invested being a much worse proxy for outcomes in a non-profit world than in the for-profit one. If I invest $1 M in Uber instead of Intel, I signal to everyone that I believe Uber is more likely than Intel to be worth more in some time horizon. If I invest $1 M in AMF instead of AWF, the signal is not clear. I might consider AMF to be more efficient than AWF but I might instead assign a higher value to human lives or simply have funds which are earmarked for GHD.
Coming back to your original post - it considers Money, Taste and Dealflow as separate components, but in our discussion you seem to propose to base the attribution primarily on Money. VC doesn't do that. Limited Partners supply the capital and get returns proportional to capital. General Partners supply taste and dealflow and get carry. The carry exists precisely because "whose money was it" and "who found the deal" are understood to be different things. A regrantor is a GP, not an LP.
So I stand by my point that it's more relevant to look at how attribution works inside firms (which also cannot rely on markets here) than to default to the amount of money invested, just because it happens to come in a familiar unit.
At least to a good approximation. Individual investors may choose to price in externalities, e.g. treat a grocery store dollar as "better" than the same dollar coming from an oil or tobacco company.