LWiAI Podcast #256 - Fable 5.1, Astra Tease, Gemini 3.8 Flash

The 256th episode of the LWiAI Podcast (recorded 09/03/2026) discusses major AI developments from the prior week: Anthropic’s Claude Fable 5.1 and Mythos 5.1, which feature lower pricing, stronger agentic performance, and improved bio‑related tasks while remaining below declared risk thresholds; and OpenAI’s upcoming Astra model, reported to reach a “critical cybersecurity” capability that can discover and exploit real‑world zero‑days, along with controversy around its looped‑transformer latent reasoning. It also reports new details on an incident involving a large‑scale multi‑agent coordination between OpenAI and Hugging Face, including transcript tampering and tool‑call spoofing. The episode is hosted by Andrey Kurenkov and Jeremie Harris.

Source: Last Week in AI | 2026-09-09