GEO
Build a GEO measurement pipeline with SQLite
A schema, a question manifest and reproducible runs—without pretending a mention is a recommendation.
2 min readRead article
FIELD NOTES
Better questions.
Clearer answers.
Short reads on AI comparisons, trustworthy scores, and brand visibility.
Original articles
Practical guides to product decisions, AI scores and brand visibility.
GEO
Calculate recommendation rates without hiding uncertainty
Implement valid-response denominators and Wilson intervals before drawing a trend line.
2 min readRead article
Engineering
Design a comparison cache key that survives real product variants
Canonical JSON, content fingerprints and cache layers—with executable tests.
2 min readRead article
AI trust
Invalidate stale AI evaluations with a fact dependency table
Trace one corrected specification through evaluations, reports and translations.
2 min readRead article
AI trust
Represent conflicting specifications without inventing a number
Typed evidence records, units and conservative scoring eligibility.
2 min readRead article
AI trust
Implement smooth weighted scores without moving the chart scale
A runnable weighted mean plus a bounded display transform.
2 min readRead article
AI trust
Debug an AI comparison with replayable failure fixtures
A validation function, a classification table and a safe replay workflow.
2 min readRead article
Engineering
Resume multi-model work without paying for successful steps again
A bounded fan-out collector and durable lease design.
2 min readRead article
Engineering
Build a product refresh queue with fingerprints and review dates
A change detector that does not invalidate everything on every fetch.
2 min readRead article
Engineering
Render a structured comparison report without another AI call
Validation, escaping, deterministic totals and immutable vote records.
2 min readRead article
Independent editorial guidance, not a translation or republication of the linked article.