Thoughts
On tech, AI, business, and what actually works.
August 22, 2026
Four AI labs spent 2026 racing to sell you a colleague made of software. The benchmarks say the best agent completes one realistic task in five, reliability hasn't improved in two years, and the standards war everyone expected never happened. The real contest is over the layer nobody demos.
August 15, 2026
A twelve-word tweet in July 2026 set off a new discipline: designing networks of AI agents instead of prompting one. Inside the evidence — Anthropic's 45-agent swarm, OpenAI's zero-hand-written-code product, the MAST failure study — and why the human's job is becoming org design plus anchors.
July 17, 2026
Mira Murati's Thinking Machines Lab shipped its first model, Inkling, and admitted in writing that it isn't the best one available. That confession is the first coherent AI business strategy built on the premise that the frontier is a loss-leader — and that the real economics live in making a good-enough model yours.
June 27, 2026
How Google turned itself from a search company into an agentic-AI company across two keynotes — I/O 2025 and I/O 2026 — and what its agents, open protocols, and quadrillion-token scale mean for business decision-makers planning their own AI bets.
June 24, 2026
SpaceX's $60 billion all-stock purchase of Cursor isn't a coding-tools story — it's Elon Musk buying the surface where software gets written, turning millions of developers' keystrokes into training fuel for his AI, and opening a second front against GitHub with a platform called Origin. What the largest startup acquisition in history means for the tools your business runs on.
May 28, 2026
The honest answer to what you get per dollar of AI spend is two numbers, not one: $3.70 for the leaders, 5.9% for the average. Which one becomes yours has almost nothing to do with the model and everything to do with preparation.