This week's major theme: the white collar reckoning. Anthropic's study measured actual AI exposure in the labor market. Computer and math occupations top the list at 94% exposure, followed by business and financial operations, legal, and management. But only 33% of that top category's workload is actually being done by AI today. That 61-point gap doesn't close on its own. It closes when leadership decides how fast to move, what to measure, and who's accountable.
Also in this issue:
- This Week - The displacement debate got data. Governance can't keep up with deployment. Business models are inverting toward outcome-based pricing. The bottleneck on AI adoption isn't capability, it's verification.
- The White Collar Reckoning - Anthropic's labor market data, Block's 40% workforce cut, Goldman Sachs automating accounting and compliance with Anthropic, and Lightspeed's framing: "AI captures labour spend, not IT budgets."
- The Wire - Thomson Reuters: only 18% of professional services firms measure AI ROI. Cursor says 35% of its own PRs come from autonomous agents. MIT Sloan: 80% of agentic AI implementation is governance, not AI. Microsoft Copilot ignored sensitivity labels twice in eight months.
- Quanta Lab - The harness matters more than the model. Why AutoGPT failed and today's agents work: planning tools, sub-agents, file system access, and 2,000-line system prompts.
- After Hours - 2001: A Space Odyssey re-read. HAL doesn't go evil. He's given contradictory instructions and his breakdown is the logical result. That felt like science fiction in 1968. It feels like a Tuesday in 2026.