AFFINE Algorithm Design
About
The Agent Foundations Field Network (AFFINE) is an AI safety organization founded by Mateusz Baginski, focused on training the next generation of superintelligence alignment researchers. The organization addresses what it views as a systematically neglected problem: despite ASI alignment being a central form of existential risk, research aimed at finding solutions is underfunded and understaffed. AFFINE's flagship program is an intensive one-month Superintelligence Alignment Seminar held at Hostacov Chateau near Prague, Czechia, running from April 28 to May 28, 2026. The seminar brings together approximately 30 participants with mentors representing leading voices across AI alignment research. Rather than pursuing novel research during the program, participants focus on acquiring deep understanding of core alignment problems through reading, peer teaching, lectures, workshops, and collaborative debate. The mentor roster includes prominent researchers: Abram Demski, Ramana Kumar, Steve Byrnes, Kaj Sotala, Kaarel Hanni, Jonas Hallgren, Cole Wyeth, Aram Ebtekar, Elliot Thornley, Linda Linsefors, Paul Rapoport, and Ouro (ex-Orthogonal). Full-time on-site mentors are Ouro and Jonas Hallgren (Equilibria Network). These mentors span institutions including MIRI, DeepMind, Anthropic, Mila, Astera Institute, Cambridge, UC Berkeley, and Harvard. AFFINE operates with two funded research focus areas: Algorithm Design and Technical Research. Algorithm Design covers the formal and algorithmic foundations relevant to alignment, while Technical Research supports broader alignment research activities. The Survival and Flourishing Fund recommended $79,000 for Algorithm Design and $165,000 for Technical Research in its 2025 S-Process. The AFFINE Fellowship is a year-long extension offered to the approximately 10 most promising seminar participants, selected for collaborative excellence. The fellowship co-locates fellows at locations where they can receive ongoing support for continued learning and research. The seminar is free of charge with accommodation, catering, and partial travel support provided. Mateusz Baginski, the founder and director, has a background in cognitive science and previously worked as a PIBBSS Fellow in 2024 under mentor Tsvi Benson-Tilsen (ex-MIRI). He transitioned from programming into technical AI safety research, focusing on theoretical and agent foundations work. The organization is fiscally sponsored by Ashgro, Inc., a 501(c)(3) public charity that provides fiscal sponsorship to AI safety projects.
Details
- Start Date
- -
- End Date
- -
- Expected Duration
- -
- Funding Raised to Date
- -
- Last Updated
- Mar 19, 2026, 6:22 PM UTC
- Created
- Mar 19, 2026, 6:22 PM UTC