MEASURED, NOT REVIEWED
AI tools, tested on real work.
One realistic task, the same input for every tool, a scoring grid fixed before the first test.
Every number is published with the date it was measured, and leads back to a kept output.
01 — THE INPUT
Same input, every tool
Each tool receives the same file, on the same day, in an order drawn at random. One pass, default settings, no second attempt.
02 — THE RULES
Rules fixed in advance
Criteria, thresholds and the conditions under which nothing gets published are written down and frozen before the first tool is opened.
03 — THE PROOF
Every number traceable
Each figure leads back to a dated run and a kept output. Inventions are counted separately from omissions.
Recently published
Meeting tools — the first corpus
Meetings come first because that is where a correct answer is easiest to write down in advance. The corpus has been written and frozen; results will appear here with their test date. Documents and paperwork follow, measured the same way — the method does not change with the subject, only the corpus does.