Writing · Topic
Research
Measured studies of the system — evaluation protocol, citation behaviour, and the paper behind them.
2 pieces
Research · August 15, 2026 · 8 min
Selection is not absorption
The most useful distinction in AI visibility, and almost nobody makes it. Being named from training data and being retrieved as a source for one answer are different phenomena with different time constants — and only one of them is work an agency can sell.
Research · August 14, 2026 · 12 min
The harness is the product
Protocol 4.0: two planes, never averaged. Judgment head-to-head — Mycel 21/21 (100%) vs a naive unscaffolded lower bound 11/21 (52%). Architecture behavior-graded 42/42 on real kernel modules; peer coding harnesses are capability-absent, not scored 0%.
Common questions
- Why do AI assistants name some brands and cite others?
- Because those are two different phenomena. Being named from training data and being retrieved as a source for one specific answer have different causes and different fixes, and almost nobody separates them.
- How do you measure whether an agent harness actually works?
- On two planes, never averaged: judgment and architecture, graded separately. Averaging them produces one number that hides which half is failing, which is the half you needed to know.
Ongoing measurement
The AI Visibility IndexHow often agencies are returned when a buyer asks an AI assistant for a recommendation — measured across real audits, updated as the sample grows.