Essay
What Should Still Be Hard
When machines can satisfy almost any present want, the harder question is which difficulties we keep on purpose, and who is in a position to keep them.
2026-10-07 · 10 min read
Read articleLong-form
Essays and field notes on constraints, coordination, and systems that can explain themselves. Each piece is published with its sources and the record of how its claims were verified.
Essay
When machines can satisfy almost any present want, the harder question is which difficulties we keep on purpose, and who is in a position to keep them.
2026-10-07 · 10 min read
Read articleEssay
“Human in the loop” names an architecture, not a control. Oversight is a safeguard only while its conditions are maintained and its detection is tested.
2026-09-29 · 13 min read
Read articleEssay
A three-layer model of what AI agents know, and why operational intent belongs in versioned artifacts rather than in agent memory.
2026-09-13 · 27 min read
Read articleEssay
After durable system structure takes over reconstruction, agent work moves toward judgment
2026-09-17 · 11 min read
Read articleEssay
Which reasoning belongs in durable structure and which must stay live: a test of standing, reuse, and revisit for the records agents and operators read.
2026-09-17 · 14 min read
Read articleEssay
As delegation matures, the human role moves from operator to governor. A better machine does not automatically inherit the authority to decide.
2026-09-22 · 12 min read
Read articleEssay
“Multi-agent” now names incompatible working arrangements. What matters is how agents are related, not how many of them are running.
2026-09-28 · 15 min read
Read articleEssay
How markets built on real wants can reshape those wants, traced through makeup, razors, gambling, pharmaceuticals, and subscriptions.
2026-09-27 · 20 min read
Read articleEssay
AI lowered the barrier to producing code in my work without lowering the standard for understanding the system. The judgment is still mine.
2026-09-25 · 8 min read
Read articleEssay
Why competitive systems can keep accelerating after everyone sees the danger—and how coordination determines whether correction is chosen or imposed.
2026-09-16 · 36 min read
Read articleField Note
Improving a constrained system relocates its bottleneck rather than removing it — and generative AI moved the constraint from production to judgment.
2026-09-13 · 13 min read
Read article