Mapping where AI could take us — and what changes the odds.

A nonpartisan research organization studying where AI could take humanity — and publishing what it finds, openly and free.

NOW O1 .190 O2 .097 O3 .236 O4 .179 O5 .034 O6 .160 O7 .105 NOW O1 .190 O2 .097 O3 .236 O4 .179 O5 .034 O6 .160 O7 .105
Seven outcome classes — thickness ∝ mass · what the seven classes are → terminal masses: model output · path geometry: illustrative · 60,000 runs · seed 42 · as of 2026-06-11

Track record

Recent work, built largely by one person with AI tooling.

  • 3,791 papers in an AI-safety literature database, with enriched metadata
  • 63,000+ citation edges mapped across that literature
  • 120 organizations analyzed, with grant flows merged
  • 101 people surveyed door-to-door on AI attitudes, published with full results and caveats