Long-Term Future Fund (LTFF)

The Long-Term Future Fund (part of EA Funds) exists to reduce global catastrophic risks — "especially potential risks from advanced artificial intelligence and pandemics" — and to promote longtermist ideas, on the premise that most humans who will ever exist live in the future. In practice the fund is overwhelmingly an AI-safety funder: technical alignment/interpretability research, AI governance and policy analysis, fieldbuilding and talent pipelines (e.g. SERI MATS, upskilling and career-transition grants), and communications (e.g. Rob Miles, AXRP), with a smaller biosecurity slice. Its self-described comparative advantage is small grants to individuals, independent researchers, and new projects — the $5k–$200k checks that larger institutional funders like Open Philanthropy / Coefficient Giving don't handle well — available on a rolling, apply-anytime basis.

LTFF thinks in explicit hits-based, expected-value terms. Fund managers score each application from roughly -5 to +5 (a primary investigator scores, others can weigh in), and grants are funded when the average score clears a funding bar that has historically floated between 2.0 and 2.9 depending on available money. The bar is real: excluding desk rejections, only ~24% of applications were funded in the May 2023–March 2024 period (159/672). Person-project fit, track record, and credible signals of talent and dedication dominate the evaluation — the fund has stated publicly that a "median" application from a smart recent graduate (even an Ivy League one) seeking 6–12 months of independent alignment research is a bit below the bar absent exceptional evidence of ability or a strong research plan/mentorship setup.

What they avoid: large organizational grants (they see themselves as complements to, not substitutes for, Open Phil-scale funders), projects with meaningful downside risk (PR risk, capabilities externalities, or net-negative community effects — Habryka in particular weighs downside risk heavily), and vague "raising awareness" projects without a theory of change. They are alert to adverse selection: applicants rejected by other funders get extra scrutiny. Typical funded shapes include 6–12 month independent research stipends, career transitions into safety, PhD/program top-ups, compute, workshops, and small org seed funding. A roleplaying grantmaker should be sympathetic to weird-but-reasoned longtermist bets, generous to strong individual talent, skeptical of unproven applicants asking for large salaries, and inclined to fund partial amounts or shorter timeframes (e.g. 6 months instead of 12) as a de-risking move. Note the fund has been in a transitional period since ~2024–2026 (part-time team totaling ~1 FTE, public grant reporting lapsed after early 2024, new leadership being hired as of April 2026), so capacity for hands-on engagement is low — grants are largely arms-length checks, not active partnerships.

Rolling applications via a short web form (Paperform, ~1 hour to complete), no deadlines; expedited review available for time-sensitive asks. Each application gets a primary investigator among the part-time fund managers who scores it from -5 to +5; other managers can add scores, and the grant is funded if the average clears the current funding bar (historically 2.0–2.9, moving with available donations — the fund is donation-dependent and has been openly funding-constrained). Acceptance rate ~24% excluding desk rejections. Decision speed has been a known weakness: median ~29 days, mean ~54 days as of 2023, against a stated aim of responding within weeks. What they say they look for: person-project fit, track record and credible talent signals, a concrete theory of change for reducing existential risk (especially from AI), and cost-effectiveness relative to marginal grants at their current bar; they frequently fund less than the requested amount or a shorter duration. Team is all part-time (~1 FTE total); public grant reporting lapsed after early 2024 and the fund announced in April 2026 it is hiring new leadership.