How the scoring works

Before a single question is asked, a scoring blueprint is drawn for the subject you named. It sets out the competencies, the sub-competencies beneath them, the keyed elements an answer must contain at each depth, and the specific wrong moves — negative keys — that count against an answer regardless of what else it said.

Blueprints are validated in code before use. Weights must normalise, keyed elements must be concrete and observable, and vague evaluative language is rejected outright. A blueprint that cannot pass validation is redrawn; if it still cannot, the sitting does not run and your credit is returned.

An examiner model asks up to 12 questions, one at a time, adapting each question to what your previous answer contained. There is no feedback, no hinting, and no easing off when you struggle — a struggling answer is followed by the question that settles what you do and do not hold.

Depth is quota'd. The examiner must reach level four (edge cases) at least twice and level five (transfer to an unfamiliar situation) at least twice. A sitting that does not reach those depths is not scored.

A second, more capable model reads the transcript three separate times without seeing its own previous passes. For each answer it records which keyed elements were present, which were absent, and which negative keys were triggered, with a verbatim quote as evidence.

An element counts as present only when at least two of the three passes agree. Where the passes disagree the spread is recorded and printed on your report, so you can see exactly where the reading was contested rather than being handed a false precision.

All arithmetic happens in code, not in a model. Sub-competency scores are the weighted share of keyed elements present, capped where a severe negative key fired. Those roll up through the competency map and the six-axis Spine into a composite, which is adjusted for the depth you sustained and expressed as a Techne score out of 1000.

The score is printed with a margin. The margin is the width the three passes actually disagreed by — two scores whose margins overlap are not distinguishable, and on the rankings they share a rank.

This is not a certification, a licence, a qualification, or a determination of competence. It is a structured record of what a specific set of answers contained on a specific day. Nothing here should be relied on as a substitute for professional assessment, accreditation, or regulatory approval.

See pricing or name a subject.