#247 - Opus 4.8, MAI, Anthropic IPO, Minimax-M3
Jun 6, 2026
More Technology podcasts
Summary
Hosts Andrei Kerenkov and Jeremy Harris discuss Anthropic’s Opus 4.8 release, highlighting its verbose personality and new dynamic workflows for complex tasks. They analyze Microsoft’s launch of the Scout agent and its in-house MAI models, noting the strategic shift toward enterprise security and custom tuning. The episode also covers Robinhood’s integration of AI trading agents and OpenAI’s new Codex tools for white-collar work, while touching on DARPA’s AI research initiatives.
Topics discussed
Intro, DARPA AI safety, and sponsor reads
Claude 3.5 Sonnet evals, sandbagging, and self-reflection
Anthropic's Computer Use and data feedback loops
OpenClaw, always-on agents, and security risks
Microsoft's new models and the 'Tier 2' lab strategy
OpenAI's IPO, valuation, and Codex productivity push
AI music tools, OpenAI's $65B raise, and economic limits
Chinese AI chips, LPU architecture, and export controls
Minimax's open source model and sparse attention
US AI policy, Meta AI hacks, and Chinese security laws
Chip export loopholes and OpenAI's Project Glasswing
NSA funding, AI extremism reports, and labeling norms
Why larger models learn more and entropy in fine-tuning
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
#258 - Opus 5.5, Sol and Luna, Muse, DeepSeek-V4.1-Flash, Xi
Oct 2, 2026
#257 - GPT 6 Astra, AI Extinction, Security Incidents
Sep 24, 2026
Listen ad-free on Castria