FIG. 0 · Writing
Writing
Things I've written about doing this work well. Everything here is free, with no email required.
FIG. 1 · White paper
WHITE PAPER · 2026
The Verifiability Gradient
Why human oversight breaks when AI output stops being checkable. Most oversight assumes someone can check the work; a growing share of AI output can't be checked, or costs as much to check as to redo. The paper sorts output into four groups by how checkable it is and says what control belongs to each. Written for the people whose job is to check AI's work, with an executive summary that stands on its own.
AVAILABLE SOONWant it the day it ships? Leave your email. No newsletter, no spam.
More is on the way.