Notes you can check
How we are building a research tool that argues back: the engineering, the classroom thinking, and the mistakes we caught in our own work. Every claim here carries its source, same as the product.
How we made an open model say Not Found instead of guessing
One row came back with four empty cells and a score of 0.18. Why abstention is harder to build than fluency, and why engagement-funded tools will not ship it.
The link works. The sentence isn't in it.
The citation resolves and the claim it supports is not in the document. What per-claim attribution looks like, worked through one CFTC enforcement figure.
What to assign when the detector is off
64.7% of Canadian teachers surveyed in grades 6-12 report no training on student AI use. A verification pass students hand in, and the rubric that marks it.