This week's major theme: the AI productivity illusion. Berkeley Haas researchers embedded with a 200-employee company for eight months and found nobody was working less — work expanded to fill the new capacity. Tiago Forte argues AI lets you "deploy vast swarms of intelligent beings to construct civilization-scale monuments to your procrastination." And Accenture found only 16% of companies moved from AI pilot to scaled deployment.
Also in this issue:
- This Week — The agent governance gap is widening (Gartner: 40% of enterprise apps will run agents by year-end, most can't control them). Block cut 40% of its workforce citing AI — Wall Street rewarded it with a 26% surge. Perplexity, Claude Code, and OpenAI all shipped remote-control agent tools. And OpenAI is hiring McKinsey to make its own product work.
- The Wire — Anthropic launches Claude Code Security and cyber stocks tank. The always-on agent race heats up. OpenAI hires the Big Four. Anthropic drops $20M on a Super PAC to fight its own industry.
- Quanta Lab — METR study: developers were 19% slower with AI assistance while self-reporting they were 20% faster. That's a 40-point perception gap. Plus why executive enthusiasm shouldn't write the press release.
- After Hours — Some Like It Hot, the musical. Billy Wilder's 1959 classic reimagined with genuine care.