
security
When Systems Fail
A runtime flaw and a breach through an AI tool's integrations: who holds the fix at each layer, and what merely buys time.
November 12, 2026Read story
The long pieces: connected stories, expanded research, and a companion notebook.

security
A runtime flaw and a breach through an AI tool's integrations: who holds the fix at each layer, and what merely buys time.
November 12, 2026Read story

agents
A note habit, a maintained wiki, standing instruction files: what persists for an agent to build on, and what upkeep earns its cost.
November 11, 2026Read story

evaluation
A leaderboard moved without an update; a proof certifies a label, not an answer. What cited AI evaluation scores prove — and leave open.
November 8, 2026Read story

agents
What you delegate is a set of boundaries, checked at named moments. What you keep is the outcome, the liability, and the bill.
November 8, 2026Read story

generative AI
Thirteen AI model launches in 26 days. What October 2026 releases actually established — and the checklist for what you can verify a month on.
October 28, 2026Read story