Coefficient Giving Is Not Funding Proposals. It Is Recruiting General Managers — $200M for Tailwind, $50M for Launchpad, EOIs Close October 2.

Most grant applications ask you to describe a project. These two ask you to describe a person — and then explain why that person should be handed a budget and told to go solve something.

Coefficient Giving, the grantmaker that operated as Open Philanthropy until its November 2025 rename, currently has two open calls running on that premise. Project Tailwind is a $200 million call for founders to launch new AI safety initiatives, on a rolling deadline. Launchpad Funds is a $50 million request for expressions of interest to seed or scale new funds in global health and wellbeing, and those expressions of interest close October 2, 2026 — nine days out.

Neither is a conventional RFP. Understanding why changes what you write.

Who Coefficient Giving is, and how much it moves

The organization was founded in 2011 out of a partnership between GiveWell's Holden Karnofsky and Elie Hassenfeld and Good Ventures, the foundation established by Dustin Moskovitz and Cari Tuna. Alexander Berger is CEO; Tuna chairs the board. Good Ventures remains the founding and largest partner, though the rename signaled a deliberate shift toward operating multi-donor funds — Patrick Collison is among the additional funders.

The scale is not marginal. More than $4 billion directed in grants as of mid-2025, with over $650 million in 2024 alone, across roughly 13 funds spanning global health, pandemic preparedness, AI risk, farm animal welfare, biosecurity, forecasting, lead exposure, economic growth, and scientific research. The organization has separately committed $1 billion to GiveWell's recommendations.

For anyone working in global health or AI policy, this is not a funder you can treat as a long shot. In several of these subfields it is the market maker.

Launchpad Funds: the DARPA program manager, imported to philanthropy

Launchpad is a $50 million call structured in two tracks:

• Track 1 (pre-seed): $200,000 to $10 million over up to 24 months, to develop the idea and strategy for a new fund.
• Track 2 (seed): $10 million to $50 million or more over up to 36 months, for funds ready to begin operating at scale.

The organizing concept is the general manager: "one person accountable for solving it — not just for making good grants in the area, but for delivering the outcome, with the authority and budget to pursue it however they think best." Coefficient points to two existing examples of the model working — the Lead Exposure Action Fund and the Strep A Vaccine Fund.

Anyone who has worked near DARPA or ARPA-H will recognize this immediately. It is the program manager model: pick a hard, bounded outcome, put one empowered person on it, give them real money and real authority, and hold them to the outcome rather than the workplan. What is unusual is seeing it in philanthropy, where the default unit of accountability is a grant agreement and a logic model rather than a named individual with a mandate.

That has a direct consequence for what a competitive EOI contains. A traditional foundation proposal spends its space on activities, outputs, and an evaluation framework. A Launchpad EOI has to spend its space on:

A specific outcome, not a theme. "Reduce childhood lead exposure" is the shape. "Improve environmental health outcomes" is not. The named precedents are both single-problem funds with a measurable end state.

Why the problem is important, neglected, and tractable. Coefficient's public selection framework is explicit about these three criteria, and it has been consistent about them for a decade. Treat them as the scoring rubric.

A named general manager with the standing to hold the mandate. If the EOI does not identify who will personally own the outcome, it is proposing a program, not a fund.

Absorptive capacity — the criterion most applicants will get wrong. Selection prioritizes funds "with potential to absorb $100m+ (and hopefully much more) from other funders over the next several years." Read that carefully: Coefficient is not looking for ideas sized to its own $50 million. It is looking for ideas that can become a durable destination for other people's capital, with Coefficient as the anchor investor rather than the permanent sole funder.

That criterion rules a lot of good work out. If your idea has a natural ceiling of $5 million a year, it may be an excellent candidate for an ordinary global health and wellbeing grant and a poor candidate for Launchpad. Applying to the wrong mechanism is a common and expensive mistake; here the mismatch is disqualifying rather than merely inefficient.

It also carries a risk worth naming out loud. A fund seeded at $10 to $50 million with an explicit mandate to raise $100 million elsewhere has significant single-funder concentration risk in year one. A candid EOI that identifies the two or three most plausible co-funders and the conditions under which they would come in reads as more credible than one that treats the co-funding as an assumption.

Project Tailwind: pre-seed money with no legal entity required

Tailwind sits under Coefficient's Navigating Transformative AI fund and is capitalized at roughly $200 million. Its tiers are venture-shaped:

• Pre-seed: $200,000 to $2 million
• Seed: $2 million to $20 million
• Scale: $20 million to $200 million

Three eligibility facts matter more than the numbers:

At the pre-seed tier, no legal entity is required. That is genuinely rare. Virtually every institutional funder demands an IRS determination letter or a fiscal sponsor agreement before a dollar moves. Tailwind is explicitly funding people who have not incorporated yet, which removes the classic chicken-and-egg problem — you cannot get funding without an entity, and you cannot justify an entity without funding.

Non-US teams are eligible. Cross-border grantmaking imposes real administrative load on the funder (equivalency determination or expenditure responsibility), and many US funders simply decline it. Coefficient does not.

Public-benefit for-profits are welcome. A PBC or public-benefit LLC can be funded. If you are on the receiving end of that, understand what it implies: a charitable funder supporting a for-profit typically does so through a program-related investment or under expenditure responsibility, which brings reporting obligations and restrictions on use of funds that a standard 501(c)(3) grant does not. Get that structure in front of counsel before you accept, not after.

Tailwind also publishes a list of concrete initiatives it would be glad to fund — rogue incident detection, agreement and verification technology, a misalignment red team, a frontier capabilities observatory, an independent AI auditing body, and talent and convening organizations, among others. Applicants can pitch their own concept or take one off the list.

The founder profile Coefficient describes is similarly explicit: strong technical ability; experience building complex hardware, software, or security-critical systems; deep research expertise in machine learning or related fields; auditing, assurance, or red-teaming experience; experience founding or scaling companies, nonprofits, programs, or events; exceptional communication with technical and non-technical audiences; or an excellent track record identifying talent.

Note the "or." That list is disjunctive, not a checklist. A skilled operator with no ML background who has scaled an organization is inside the target profile.

What changes about the application craft

Both calls reward a document most grant writers are not practiced at producing. Concretely:

Lead with the person and the mandate, not the need statement. In a normal proposal, paragraph one establishes the problem. Here, the problem is largely stipulated — the funder picked the domain. What is unknown is whether you are the right person to own it.

Write a theory of change that terminates in an outcome, not a set of deliverables. "We will convene twelve stakeholders and publish four reports" is a workplan. "Blood lead levels in the three target countries fall measurably by 2030, and here is the causal chain" is a mandate.

Be explicit about what you would do with each tier. Both programs are tiered. A strong application states what it can accomplish at pre-seed, what changes at seed, and what would have to be true for the scale tier to make sense. That structure also gives the funder an easy way to say yes at a smaller number instead of no at a large one.

Say what would make you wrong. Coefficient's culture runs on explicit reasoning about uncertainty. A section naming the two assumptions that would sink the plan, and what early evidence would test them, reads as competence rather than weakness.

Budget for the operating entity, not just the program. Under the GM model you are proposing to build an institution. Recruiting, ops, legal, and a grants function are part of the ask, and understating them is a credibility problem, not a cost saving.

The distinction between these calls and the broader foundation payout picture is worth sitting with. While the median billion-dollar foundation distributed 5.1 percent last year, this is a funder actively trying to build new vehicles specifically because it believes it cannot deploy fast enough through the ones that already exist. That is an unusual problem to have, and it is temporarily an applicant's advantage.

Launchpad expressions of interest are due October 2, 2026. Tailwind is rolling, which in practice means the earliest strong applications get the most attention and the least competition. If you have been carrying an idea that never had a funder shaped like this, the constraint on submitting is now nine days of writing — and tools like Granted exist to compress exactly that gap between a formed idea and a document a program officer can act on.