OpenAI's Codex engineer keeps a doc of wrong AI predictions — so she only plans 2-3 months ahead
victor_explore · x · 2026-09-30
Tara Seshan, who works on Codex at OpenAI, keeps a doc of her own wrong AI predictions to remind herself she can't see years out — so she only plans 2-3 months ahead. Build for today's models and you fall behind, but anything past 90 days is a guess.
The reblogger adds: a doc of your own wrong calls is the cheapest eval you can run on yourself; if the people shipping Codex won't plan past 90 days, a 12-month AI roadmap is a guess with a Gantt chart.
More from AGI Musings
- Brundage: unheeded warnings should be the baseline assumption for all AI incidents — Miles_Brundage · 2026-09-30
- OpenAI exec: deciphering ancient scripts is what superintelligence excites me most — BorisMPower · 2026-09-30
- We're entering a world of Jevons paradox as AI efficiency gains fuel more usage — realsohamparekh · 2026-09-30
- OpenAI reportedly replaced research interns, and junior roles may be next — cneuralnetwork · 2026-09-30
- SchmidhuberAI: AI proofs will be like compiler output — individual researchers' value gone in a year — ethanCaballero · 2026-09-30
- Developer: Always-On AI Agents Like Muse and Grok Bot Must Be Open Source — yacineMTB · 2026-09-30