Research
Research: what the logs actually show
Short papers from my own practice. Each one makes one claim, shows the numbers behind it, and says what would change my mind. They are written for people who run AI-assisted work.
What 1,495 transcripts reveal about how AI models actually differ
A personal AI harness with 100+ behavioral rules and zero measurements
AI made production nearly free. The last mile got more expensive.
Working with AI is not prompting. It is building an operating system, and the receipts are in the logs.
72% of an AI operation ran unattended, and that was the easy part to explain
You cannot state where your hours go. Your logs can.
The only honest dataset about your AI operation, and the two ways you are losing it
If you measure AI corrections from chat text, you are measuring the minority signal
In agentic work, completion is testimony, and testimony is the cheapest thing a model produces
Every AI memory product stores what happened. None of them store what you corrected.
The corpus says model choice is procurement, not preference
Next step
Want this measured on your own company? Start with a free 20-minute diagnosis, or see the Time Audit.
Last updated: