Academica runs internal research programmes as designed experiments against live systems. This is where they are published in full: the hypotheses as they were written before the run, the reports as they came out, and the ledger of what every experiment settled — including the ones that failed and the ones that were blocked.
There is no sign-in. Nothing here is summarised for an audience; the figures are compiled from the runs' own results files, so the site cannot disagree with the data it came from.
PROGRAMME 01
Jev — an observation client, answered by experiment
108 experiments in 7 groups against 38,194 observation records. 55 confirmed, 29 falsified, 10 mixed, 1 inconclusive, 13 blocked.
108 experiments · 38,194 records · 45,861,280 input tokens · $1.9262 at $0.042 per million.
Compiled from dependency-docs-reference/jev/notes/12-experiments on 2026-09-19.