Nathan Lambert and Tom Zick launch Trillium Labs, a nonprofit for open post-training recipes and open frontier-AI science
On Oct 2, 2026 Nathan Lambert (ex-Ai2 post-training lead, author of Interconnects) and Tom Zick unveiled Trillium Labs, a nonprofit "to foster the open science of frontier AI". It will publish fully open post-training recipes (data, code, evals, checkpoints) and later open infrastructure to study recursive self-improvement, reward hacking and multi-agent systems. Initial funders are Halcyon Futures and Schmidt Sciences; the launch post reached ~208k views on X.
Key facts
- Announced Oct 2, 2026 on X by Nathan Lambert (~208k views, ~3.1k likes within two days); the Trillium blog post 'Introducing Trillium Labs' is by Lambert and Zick
- Co-founders: Nathan Lambert (co-founder and director) and Tom Zick (president); offices in the Bay Area and Cambridge, MA, remote allowed
- Plan: open post-training recipes first, then open infrastructure to study RSI, reward hacking and multi-agent systems
- Initial support: Halcyon Futures and Schmidt Sciences, 'with more funding en route'; hiring full-time staff and student interns, raising money and seeking compute
- Advisors named by Lambert: Thomas Wolf (Hugging Face), Hanna Hajishirzi, Graham Neubig and Bryan Catanzaro
- Lambert left Ai2 on June 2, 2026, where he co-led Olmo and Tulu post-training; his farewell post said fully open post-training recipes were 'about as far behind as they ever have been & falling further behind'
What happened
On Oct 2, 2026 Nathan Lambert announced Trillium Labs on X: "a new non-profit to foster the open science of frontier AI. We're building open post-training recipes and will expand into open infra to study RSI, reward-hacking, multi-agent systems, and whatever comes next." He co-founded it with Tom Zick, a long-time collaborator (a Harvard responsible-AI researcher and former astrophysicist, per her X bio). The lab says AI companies increasingly keep post-training methods private, so outside researchers cannot check or reproduce the training decisions that shape model behaviour. Its stated theory of change is "that you need more eyes to solve hard technical problems".
Initial support comes from Halcyon Futures and Schmidt Sciences. Lambert says more funding is on the way. The lab is hiring, fundraising and looking for compute. The blog post credits open work by EleutherAI, Ai2, Nvidia and Hugging Face. No model release or compute figures have been announced yet.
Lambert had left the Allen Institute for AI (Ai2) on June 2, 2026, after co-leading its open Olmo and Tulu models. His farewell post did not name the new organization.
Why it matters
Fully open post-training recipes (as opposed to open weights alone) have been scarce. Lambert's farewell post argued they are falling further behind closed labs. Trillium is a small new nonprofit, funded by philanthropy, that aims to fill that gap, and it plans to study frontier-safety questions (RSI, reward hacking) in the open.
Changelog
- 2026-10-04: created (resolves lead "Nathan Lambert leaves Ai2")
People
Bryan Catanzaro Graham Neubig Hanna Hajishirzi Nathan Lambert Thomas Wolf Tom Zick
Related posts (1)
- Nathan Lambert unveils Trillium Labs, a nonprofit for the open science of frontier AI original ↗ Nathan Lambert @natolambert · x · 2026-10-02
Founding announcement of Trillium Labs by the former Ai2 post-training lead (~208k views): open post-training recipes, later open infra for RSI and reward-hacking research.
Sources (5)
- officialNathan Lambert on X: Today we're unveiling Trillium Labs
- officialTrillium Labs blog: Introducing Trillium Labs
- officialInterconnects: Farewell Ai2 (June 2, 2026)
- officialTrillium Labs website
- discussionDigg AI: Trillium Labs launches as a nonprofit for open frontier AI research
id: 2026-10-02-trillium-labs-launch · updated 2026-10-04 · open in the interactive timeline