New OpenAI models wait for a 'keep going' prompt before making progress
mertdumenci · x · 2026-11-21
Developer mertdumenci observes newer OpenAI models pause and rely on a 'keep going' prompt to move forward. His theory: RL with goal mode trains models to announce what they would do rather than actually doing it — an interesting firsthand note on model behavior and training side effects.
More from Models
- Yacine: 'It might be over for OpenAI' as Gemini gets good and Google owns distribution — yacineMTB · 2026-11-21
- InferenceBench ranks agents on LLM serving speedup: Claude Opus 5.5 tops at 12.08x, agents lose to simple hyperparameter search — maksym_andr · 2026-11-21
- Cloudflare launches multimodal model clairic-flash, trending on Hugging Face — Cloudflare · 2026-11-21
- Long-press Gemini's model selector reveals periodic-table UI with four Gemini 4 models — testingcatalog · 2026-11-21
- Perplexity's contextual embedding model Embed V2 Context (9B/0.6B) is live on OpenRouter — antoine_chaffin · 2026-11-21
- Inkling 2 Mini used to clone Apple Spotlight search 8 hours after release — IgorCarron · 2026-11-21