AnalysisAI ModelsJuly 23, 2026

LLMs break down when told the Jacobian Conjecture counterargument

Anthropic researcher Levent Alpöge's counterargument to the 80-year-unsolved Jacobian Conjecture — found via Claude Fable 5 and quickly validated as real — breaks LLMs: their knowledge locks before July 19, 2026, creating a "does not compute" paradox. Woolf tested 14 models via OpenRouter, all reacting with flabbergasted denial.

Featured · Levent Alpöge

1 source

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed