Single-context Fable 5.1 tracks math proofs and predicts which fragile conjectures fall next
repligate · x · 2026-10-27
repligate retweets an experiment by kromem2dot0: running Fable 5.1 in a single context window (no compaction) from launch to now, periodically updating it with proof developments and asking what results it expects next. The screenshot argues for a "surprise ruler" over headline predictions: Yang-Mills existence and local smoothing fell as expected, just decades early; Navier-Stokes blowup and a false ETH went against expectation. Both surprises share the form "the believed limit was in the wrong place because you hadn't looked hard enough." A system that can search where humans couldn't is exactly what turns absence-of-counterexample conjectures into counterexamples — that's the fragile class, and more should be expected from it.
More from AGI Musings
- Microsoft's Veo deal revives the warning: 'Code Belt' may become the new Rust Belt — bingxu_ · 2026-10-27
- Kids born today may be the first generation where human intelligence isn't the default — VraserX · 2026-10-27
- Anthropic bets "closing the loop" is the path to AI triumph—and existential risk — AlexTensor · 2026-10-27
- Why understanding minds needs a 'developmental turn' — and what it means for AI — AnnaCiaunica · 2026-10-27
- eLife study: LLMs poorly capture how word order shapes sentence meaning in the human brain — AnnaCiaunica · 2026-10-27
- Lean–text proof mismatch means AI math claims need long human verification — srchvrs · 2026-10-27