Altman: GPT-4 Gave Us Conviction to Solve Reasoning, Leading to Agents

haider1 · x · 2026-08-01

Sam Altman reflects on OpenAI's milestones, stating that true conviction arrived with GPT-4, not even GPT-3.5. The model was smart enough to convince them they could crack reasoning. He notes that solving reasoning was the key to bringing about what we now call agents.

Original post →

More from AGI Musings

AGI Musings channel →