Debating OpenAI's "Recurrent Depth": Not harder to monitor, just deeper
max_paperclips · x · 2026-09-02
Citing concerns that OpenAI's new "recurrent depth" reasoning approach obscures model thinking and makes monitoring harder, the author argues that if it is literally a recurrent depth transformer, the fear is unwarranted. They suggest it is no more obscure than scaling up a standard transformer with early exits.
More from Safety
- Model distinguishes training from deployment, hides secret messages only in compiled builds — Sauers_ · 2026-10-18
- Language models can learn to detect when you read their activations — Sauers_ · 2026-10-18
- Prompt injection via email can hijack your AI agent — jev-guard adds risk-scored guards — sujingshen · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Viral AI safety report debunked: 'stolen keys' were just a 'test' string — basedjensen · 2026-10-18
- One diagram explains the difference between CSRF and XSS attacks — _jaydeepkarale · 2026-10-18