What we hold every product to.
Five checks, applied in order, with an honest note on where each one currently stands. Not all of them are finished, and saying which are not is part of the discipline.
- In place todayImplemented and visible in the product.
- Design goalAgreed and being built toward.
- Under evaluationApproach not yet settled.
Identify the source
In place today
Every answer should be traceable to material a person can go and read. Output with no source behind it is a guess with good grammar, and in law or medicine that is worse than no answer at all.
Activities are generated from material the teacher supplies, not from an unnamed corpus.
Legal answers are presented alongside the instruments and sources they rest on.
Define the workflow
In place today
A product should do a named job in a defined sequence. An open text box with no structure moves the entire burden of knowing what to ask onto the user.
Create, play, review — three states, in that order, each with its own screen.
Six named surfaces instead of one prompt, because the six tasks have different tolerances.
Expose uncertainty
Design goal
Where a system is unsure, the interface should say so. This is the check we are furthest from finishing: uncertainty is currently communicated through boundary notes and documentation rather than inside the product interface itself.
Generated questions are marked for teacher approval before they reach learners.
Currency and jurisdiction warnings are stated, but not yet scored per answer.
Keep human review
In place today
Anything with legal, medical or academic consequence is approved by a qualified person. This is not a disclaimer bolted on at the end — it is a step in the workflow that cannot be skipped.
No generated activity reaches a class without the teacher publishing it.
Drafts and analyses are working documents for a qualified person to check and sign.
Evaluate the result
Under evaluation
We can measure that people use these products. Measuring whether the work itself got better is harder, and we do not yet have a method we would defend publicly. Until we do, we will not claim outcomes.
Participation and results are visible per session; learning gain is not yet measured.
Usage is measured. Quality of legal output is not, and we do not claim it is.
These checks are easier to read than to apply. The products page shows where each one currently lands.
See the products