Everything the lab publishes lives here: measurements taken from our own corpus, method notes, replications and results that did not work, and arguments about where commercial behavioural data is going. Findings and papers report what was measured and attach the sample size. Theses argue a position and say so on the first line. Notes are shorter pieces of either kind. We publish when there is something to publish, at most weekly, and never on a schedule that outruns the evidence.
What the corpus counts as of 2026-09-29 do and do not establish about whether behavioural evidence of competence could be issued to a person and verified by a later employer.
What a count of represented frameworks does and does not establish about evidential support, and which admission criteria the available evidence leaves unreported.
Roughly a quarter of assessment observations return no grade, and the three reasons behind that are structurally different measurements that should not be pooled.
A table whose name asserts human authorship holds 1,145 independent AI dual-rater re-grades and 9 rows recorded by a human operator, with no cost figure available.
Why the published audit literature and the EU AI Act's internal-control route together leave AI scoring of people at work without an independent check, and what one would require.
Why per-person payment for behaviour that trains a model is currently blocked at attribution, entitlement and valuation, and what published work would have to settle first.
