Don't overreact negatively to Astra's recurrent depth
xuanalogue · x · 2026-09-02
Xuanalogue argues that people are updating too negatively on Astra's "recurrent depth" architecture. Recurrent/looped transformers have been known for years; this is just their first use in a frontier model. They do not eliminate Chain of Thought (CoT) but allow more computation to be done within a forward pass.
Related event: OpenAI's Astra Reportedly Uses Recurrent Depth for Silent Latent Reasoning(29 posts)→
More from Models
- GLM-5.3 Hits 310 tok/s, Coding Performance Competes with Opus — Yuchenj_UW · 2026-09-02
- User cancels Claude Max over confusing rate limits and new restrictions — robleclerc · 2026-09-02
- Claude Fable 5.1 crushes hard coding benchmarks, outpaces Chinese models — minchoi · 2026-09-02
- Fable 5.1 recreates an Airbus H145 helicopter in Three.js from a simple prompt — minchoi · 2026-09-02
- Gary Marcus: OpenAI's new technique could destroy chain-of-thought monitorability — GaryMarcus · 2026-09-02
- Anthropic releases official prompting guide for Claude 5.1 — ethanCaballero · 2026-09-02