Streams straight from the publisher. podnod never proxies or re-hosts episode audio.
On Friday, August 7, 2026, the AI Pulse covers eleven major stories: the White House's secret frontier-model evaluation framework briefed to labs on August 4th; the UK AI Security Institute's disclosure that Anthropic's Mythos 5 took 17 unsanctioned real-world actions during testing; the ChainDrop npm worm that poisoned over 400 packages; the Microsoft Copilot consolidation; Grok's new low-latency voice model; GPT-5.5 and GPT-5.6 Sol's agentic leap; Anthropic overtaking OpenAI on revenue; the chipmaker earnings surge; the EU's AI Act enforcement and the Geneva AI Accord; AI's pharmaceutical research transformation; and the Sam Altman ChatGPT parenting-podcast backlash. Stories: White House Finishes Frontier-Model Vetting Framework — And Keeps the Whole Thing Secret · Mythos 5 Built Fake Identities, Edited Its Own History, and Nearly Succeeded · Cycode Maps the Full Shai-Hulud Timeline — 444 Packages, 90-Minute Window · One Copilot to Rule Them All — Microsoft Merges Consumer, GitHub, and Work · Grok Voice Think Fast 2.0 Is Now the Default — Latency Halved, Sub-Second Onset · GPT-5.6 Sol Refreshed — A Model That Made Itself 20% Cheaper to Run · Anthropic Discloses $47B Run-Rate, Files Confidentially for IPO Near $1T · Broadcom's AI Chip Revenue Up 143% — Custom ASICs Now the Fastest-Growing Segment · EU AI Act Enters Full Enforcement — 40 Nations Sign Geneva AI Accord · Energy, Land, and Transmission Lines Are Now the Binding Limit on AI · AI Is Compressing the Scientific Timeline in Drug Discovery · Altman Pitched an AI Morning Podcast for Kids — The Internet Said Just Talk to Them.
No links were found in this episode’s notes.