OpenAI Astra rumored to be a looped transformer, rasbt steps in to debunk the hype
prdeepakbabu · x · 2026-09-03
A new The Information report claims OpenAI's rumored Astra model uses a "recurrent depth / looped transformer" architecture, sparking discussion. Sebastian Raschka (rasbt) published a measured debunk, noting he shared analysis of similar architectures two months ago and urging skepticism about closed-lab rumors. prdeepakbabu links related material in response. A technical discussion of the rumor rather than confirmation.
Related event: OpenAI's Astra Reportedly Reasons in Latent Space via Recurrent Depth(88 posts)→
More from Models
- ChatGPT helps resolve 30-year-old stable forking conjecture in logic — Dr_Singularity · 2026-09-03
- Users praise Gemini 3.8 Flash: "coding is fun again" — xennygrimmato_ · 2026-09-03
- Grok Bot's Token Limits Frustrate Users; Subscriptions With All-You-Can-Eat Tiers May Win — NickPassig · 2026-09-03
- Author Burkov says Anthropic ignored his refund requests, so he cobbles together Codex and Grok — burkov · 2026-09-03
- Google launches Gemini 3.8 Flash Cyber security model alongside Fairwind Program for defenders — GoogleAI · 2026-09-03
- Matt Shumer: slow AI releases aren't a wall — safety clearance is the bottleneck, wave of frontier models imminent — mattshumer_ · 2026-09-03