Logbook
We show how we build, mistakes included: architecture choices, evaluation methods, analysed refusals to answer. Each note sets out what was tried, what was measured where measurements exist, and what was kept. Examples rely on public or anonymised corpora; no client data ever leaves its perimeter to appear here.
Note 1 · Scoping ·
Behind the request “we want to do AI” lie very different projects. Insufficient scoping is paid for later, in delays and budget. Here are the questions we settle before writing a line of code.
Read the noteNote 2 · Expert selection ·
A CV describes what someone has done, in contexts that no longer exist. A skill profile answers a different question: what can this expert do for you, now, in your context?
Read the noteNote 3 · Evaluation ·
For this site's demonstration, we queried Knowledge AI on an extract of sixteen articles of the French Labour Code. One question was meant to be refused, two were meant to be answered. The result is more instructive than expected: it shows what a refusal proves, what it does not, and how to check it.
Read the noteComponents under development, with their actual state and measurements, are tracked in Build in Public.
Open Build in PublicTell us about your context: we will tell you what our experience allows us to conclude, and what it does not.
