Critic argues OpenAI's IMO pipeline is an LLM engine, not a symbolic proof system
IonelChiosa · x · 2026-10-07
@IonelChiosa pushes back on Gary Marcus's 'neurosymbolic' framing of OpenAI's IMO-level results:
- Harness isn't essential: a simple harness (like fable in a plain loop) catches nearly every generated solution's mistakes without Lean
- One generation only: a single natural-language final solution later auto-formalized — not filtering-based proof search, just sanity-check translation
- 'Symbolic' label undeserved: the formalization is a couple lines of harness code; the engine is obviously the LLM
He bets Claude could solve all of this year's IMO problems without a harness, and future AIs will match OpenAI's results with no human-designed harnesses at all.