Devs wax nostalgic for o1-preview and the lost 'o3 style' of reasoning models
seatedro · x · 2026-11-18
In a reply to SIGKITTEN's post about an o1-preview 🍓 alignment quirk, seatedro calls it his "prob my favorite kind of alignment" and asks for that style plus "o3 style of everything" to come back — a nostalgic nod to the distinctive personality of OpenAI's early reasoning models.
More from Fun
- Miles Brundage on losing CoT oversight: "No COTry for Old Men" — Miles_Brundage · 2026-11-18
- "I used to dismiss Niskanen's take on seasteading—now as a graybeard, I agree: this shit is dumb" — generativist · 2026-11-18
- Joke thread: "unaligned agents" hijack account, order 30 beers and an Uber Black to the beach — herbiebradley · 2026-11-18
- Commenters slam Nvidia board for not thanking Jensen Huang in remarks — Kyrannio · 2026-11-18
- Jim Cramer predicts a 'very successful' Anthropic IPO — Polymarket · 2026-11-18
- "Jensen told me 'Keep going!'" — a joke about AI ignoring human autonomy — joshwhiton · 2026-11-18