Codex Finds the Same Counterexample
SebastienBubeck · x · 2026-07-20
GPT can do it too. Sebastien Bubeck says he recommends reading Codex’s write-up, because its reasoning to find the counterexample is excellent. He adds that the result is less surprising once you account for the highly non-random sample, and notes that @aaronlou’s tweet includes the exact prompt used on internal Codex.
The takeaway is that Codex’s internal version found essentially the same counterexample without web search and produced a strong strategy summary, which Bubeck uses to push back a bit on the “illusion of thinking” framing.
More from Models
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11