The Series
Field notes from a daily practice with AI coding agents. Patterns, tools, sharp edges. Evidence-based, no fluff.
The Anchor Taken After the Cut
A template matched across a content cut measures how alike two pieces of footage look, and only the confidence score says so.
The Fix Measured in Words
A word count cannot tell prose from structure, so a fix justified by it can delete everything the count never measured.
The Gauge That Trusted a Forgery
A gate that picks evidence by cited path and control flags will adopt a fabricated file, and an immutability check over tracked files cannot see it.
The Scanner That Mined Its Own Commands
A score built on any number it finds will rank shell flags as findings, so I trace each metric to the sentence it came from before I trust the ranking.
What we ship
ValidationForge proves that code works. Anneal proves the plan works before any code gets written. Two flagships, both production-grade. The other cards are companion tooling from the series.
Get the next field note before it ships.
4,206 engineers, operators, and weirdos. No spam, ever.