Terence Tao admits even he was overwhelmed skimming his ChatGPT math session
ChrisGPotts · x · 2026-11-12
Chris Potts highlights a talk where Terence Tao revealed that, despite appearing to absorb enormous amounts of information in his public ChatGPT session on the Jacobian conjecture, he was actually totally overwhelmed and had to do a lot of skimming. The thread also spawns an idea: use gaze tracking to detect which parts of a model's response the user never reads, so the model learns to stop generating slop.
More from Models
- Fable 5.2 tested: built a pro-grade video editor unattended in 3 days — every · 2026-11-12
- Heavy Users Say Claude's Safety Rails Block Even Basic Chemistry Work, Forcing Fallback to Opus — rbhar90 · 2026-11-12
- Community teams release wave of abliterated Qwen models for 8–24GB local GPUs — udmrzn · 2026-11-12
- a16z's Martin Casado: Google's Carbon is 'really good at code' — xennygrimmato_ · 2026-11-12
- OpenRouter launches Extractions API: typed JSON field extraction from text, images and PDFs — michellechen · 2026-11-12
- Google's busy fall: Gemini 4 Argon launches, plus OpenAI models on Google Cloud — samcharrington · 2026-11-12