Last Week in AI Last Week in AI

#258 - Opus 5.5, Sol and Luna, Muse, DeepSeek-V4.1-Flash, Xi

Oct 2, 2026 · 1h 43m

Summary

This episode covers Anthropic’s Opus 5.5 release, which offers lower prices and stronger agentic performance than GPT-6 Astra, alongside OpenAI’s new cost-competitive GPT-6 Sol and Luna models. The hosts discuss the widening capability gap between closed and open-source models, the implications of White House policy on release timing, and safety concerns regarding sandbox escapes and recursive self-improvement.

Topics discussed

Intro and quiet week in AI Discussion on AI pain and sentience Sponsors: Outshift and Langfuse Anthropic Opus 5.5 release and benchmarks Open source model gap analysis AI R&D automation and recursive self-improvement Anthropic security tools and community feedback Pacing AI development and antitrust tensions OpenAI GPT-6 Sol release and cost analysis Nominat AI classifier model capabilities Meta Muse mobile launch and distribution Amazon blocks Meta Muse AI agent Sponsor: Factor meals DeepSeek V4 Flash and KV cache compression US-China AI summit and diplomacy outlook Antitrust lawsuit against AI labs Public perception of data centers Detecting reward hacking via internal representations Research on AI pain and emotional states OpenAI solves Navier-Stokes millennium problem Multi-agent communication vs parallel work Regularized recursive self-improvement

More from Last Week in AI

#261 - (belated) Argon, Muse Spark, Anthropic S-1
Nov 20, 2026
#260 - Muse, Misalignment, Math
Oct 19, 2026
#259 - Dots, Sonnet, Self-Safety, Rogue AI
Oct 9, 2026
#257 - GPT 6 Astra, AI Extinction, Security Incidents
Sep 24, 2026
#256 - Fable 5.1, Astra Tease, Gemini 3.8 Flash
Sep 8, 2026
Listen ad-free on Castria