UofAi

Pilot evidence

The evidence, as we collect it

Measured

Observed directly in a defined sample. We publish the method, retrieval date, and sample size.

Estimated

A bounded estimate based on disclosed inputs. It is never presented as an observed outcome.

Illustrative

A worked example that shows the method. It is not learner evidence or a product-performance claim.

Illustrative artifact · not learner evidence

From a vague request to a defended research brief

Before the Rep

“Summarize how competitors price their products.”

No audience, decision, source boundary, or standard for a trustworthy result.

After Frame → Ground → Direct → Critique → Capture

A decision-ready brief covering five named competitors, with every pricing claim checked against a primary source and one unsupported generalization removed.

Illustrative rubric: framing 3.5 · tool leverage 3.0 · judgment & verification 3.5 · output quality 3.0 · transferability 3.0.

This is a model demonstration of the assessment method. It is not a testimonial, pilot result, or claim about an actual learner.

Pilot evidence register

As of 30 August 2026, consented design-partner evidence is still being collected. We will publish each qualifying entry here after written consent and review, with its sample size, source, collection date, and evidence classification. We will not substitute invented quotes or unlabelled model examples.