OpenAI's Astra uses recurrent depth to hide reasoning, but experts say it's a minor tweak
dotey · x · 2026-09-06
The Information reports OpenAI's Astra uses "recurrent depth" (looped transformer), letting the model repeatedly process the same text for higher quality while hiding part or all of its chain of thought, making internal reasoning hard to inspect.
- Insiders say OpenAI deliberately limited the technique so Astra still produces human-readable CoT for monitoring; the official blog also promised "additional chain-of-thought monitoring" at launch.
- Industry researchers, including the U.K. AI Security Institute, worry other developers adopting the technique without such restraint could produce unmonitorable "runaway AI."
- Architectures expert Sebastian Raschka counters that looped transformers aren't new nor very important: "Astra may be a great model, but it likely has little to do with it — just a minor architectural tweak."
Related event: OpenAI's Astra Reportedly Uses Recurrent Depth, Raising AI Safety Concerns(7 posts)→
More from Models
- ICML paper: finite context windows impose impossibility results for adversarial regret — burny_tech · 2026-11-02
- SenseTime open-sources SenseNova-Loop-9B: weight-shared 56-layer loop, 256K context, Apache 2.0 — AdinaYakup · 2026-11-02
- User says Gemini excels at long-form fanfic, producing stories up to 29 minutes of reading — Ok-Jello3739 · 2026-11-02
- Sam Altman calls tokens 'a kind of fake unit' that doesn't measure business value — rohanpaul_ai · 2026-11-02
- 'You Can Feel the Gradient Slopes': Dev Contrasts Benchmaxxed RL Models vs General Ones — theteknosaur · 2026-11-02
- LightOnOCR-3 Launches: Tops FRBench, Adds Layout Structure and Chart Extraction — IgorCarron · 2026-11-02