Thinking
Essays, when an idea earns its length.
9 July 2026·5 min read
Autonomy is earned, not assigned
Long AI chains fail quietly, and the errors compound. The teams getting the most out of AI work in spurts — and promote a workflow to full automation only after it has survived enough human checkpoints.
14 June 2026·4 min read
A Fable of Two Builds
Two months ago the same biotics subject produced a static page. Now it's a playable 3D world — built by a model the US government ordered pulled days after it launched.
20 May 2026·5 min read
What gets revalued when the legible work compresses
A category of work has been quietly accumulating names for a decade. None of them quite fit. Agentic AI is about to make the gap impossible to ignore.
21 April 2026·12 min read
The $150 Stack: Asymmetric Opportunity, Asymmetric Risk
AI has collapsed the cost to build. Not the cost to run a business. The distinction matters more than most people currently appreciate — and it has different implications for builders, enterprises, and the competitive landscape between them.
16 April 2026·9 min read
The Bundled AI Bet: Why Enterprises Are Losing a Race They Think They're Running
Enterprise AI has a bundled vs best-of-breed problem — the same pattern that killed Siebel. Arena data, Recon Analytics adoption numbers, and the iteration compound nobody is measuring.
14 April 2026·5 min read
What seven years in a lab taught me about building products
The scientific method — hypothesis, test, measure, revise — is the most underused operating system in digital product development. Here's why it matters.
8 April 2026·5 min read
Claude vs Copilot: Same Prompt, Two Outputs, No Cleanup
A controlled comparison — identical prompt, cold start, no post-processing. What two AI tools produce when you ask them to build the same thing, and what the gap looks like after one iteration.