CERTESIAN RESEARCH
Build it.
Question it.
Keep the evidence.
Our journal of engineering decisions, useful failures and the tests that change what we build.
RESEARCH #001
An AI clinical examiner you can audit.
Earlier product experiments on citations, repeatability and distinct report policies.
Recorded product research RESEARCH #002The time was right. The explanation was invented.
204 Gemini calls revealed invented factual explanations in witness roleplay.
Synthetic actor & evaluator probes RESEARCH #003One evidence layer. Different assessment rules.
A shared contract, source-compatible adapters and a harder mixed-outcome evaluation suite.
Implementation & benchmarkTRY THE CURRENT IMPLEMENTATION
Open the workbench Examine a judgement yourself.
Load a synthetic transcript, run the evaluator, and inspect its cited evidence.
These are engineering studies. Small synthetic suites are not clinical validation, readiness prediction or evidence of hiring suitability.