agi.fyi Launches as the 'Wirecutter for AI Risk Content' With Five Curated Guides
KatjaGrace · x · 2026-09-18
Researchers in AI safety launched agi.fyi, a 'Wirecutter for AI risk content' that curates the best reading, watching, and listening material for newcomers. It debuts with five guides: the best articles on how AI could actually endanger humanity, top AI-risk videos, government policy plans with concrete enforcement, a primer on alignment research progress, and a reading list on the recent Hugging Face incident. Curation relies on experienced contributors plus surveys of full-time researchers, starting from newcomer questions rather than intra-field debates. Endorsements include Zach Stein-Perlman, Daniel Kokotajlo, and Buck Shlegeris.
More from AGI Musings
- Anthropic publishes three transparency metrics: AI-driven R&D, agent oversight, compute allocation — repligate · 2026-09-18
- The containment culture paradox: those who could constrain agents wouldn't fit AI labs — repligate · 2026-09-18
- Commentary: alignment norms aren't a natural convergence, and Anthropic bears blame — ryunuck · 2026-09-18
- Palantir CEO Karp: closed-model token discounts are really about harvesting your IP — rohanpaul_ai · 2026-09-18
- The Spectator: Silicon Valley rumors say some Anthropic engineers have begun 'worshipping' Claude as a god — Polymarket · 2026-09-18
- 1a3orn: if AI deserves rights, the time to grant them is during training — 1a3orn · 2026-09-18