VontosEvidence of Learning

The journal

Written when something real happens.

No content calendar. Entries appear when the work produces something true — a regulation read closely, a doctrine argued, a lesson from real classrooms once the schools involved say it may be told.

2026
AUG 14DOCTRINE

Why no AI grades anyone here.

A named person attests every rating in Vontos. What we let machines do, what we never let them do, and why the line sits where it sits — including the part where the honest answer is stricter than the regulation requires.

CARLO BELTRAN · SOURCES: PRODUCT DOCTRINE · EU AI ACT · NPC 2022-04
CORRECTION · AUG 12

This entry's draft implied human attestation exempts a system from AI-regulation risk tiers. It does not — attestation is right on the merits, not a classification escape. Rather than delete the sentence, we published it struck through, with the correction appended. That is how this journal treats being wrong.

One entry is not a slow start. It is the honest size of what we know so far.