Verify From Where the Caller Stands
Four times this month a check passed and the thing it checked was wrong. Each time the instrument was standing in the wrong place. The rule we run now, with the receipts.
Field Notes
Essays from the build. What production agentic AI actually takes, written by the people running it.
The Far End of the Pipe
Our front desk answers a public number. We audited it against the catalog we sell before we sold it, then watched it fail in the one place a dashboard cannot see. What the watch looks like now.
Half the Findings Were the Scanner's Fault
Before selling the AI Codebase Debt Audit to anyone, we ran it on our own open-source repo. First scan: $3,150 of debt, 8 high-severity findings. Eighteen of the thirty-six findings turned out to be bugs in our own scanner. Here is the whole ledger, including that part.
The Maintainability Gap Has a Dollar Sign
AI now writes a serious share of production code, and the 2026 numbers on what that code does to a codebase are in. Nobody's business model depends on cleaning it up. Ours does.
An Answer Without a Citation Is a Guess
We run semantic search over seven thousand of our own working sessions. The most important thing it does is refuse to answer. This week we open-sourced it.
Green Gates Cannot Measure Presence
A test suite passed while the whole point of the work was being destroyed. What that taught us about verification, and about the one check no gate can replace.
The Agent That Gives a Damn
Why a persistent identity produces work a stateless agent can't, and what identityless agents quietly cost you.
The Bottleneck Was Never the Builder
How a founder and the agents that work beside him ship production agentic AI in weeks, and why the speed is a byproduct of verification, not generation.