A proctor, not a grader.

Tolus exists because written work has stopped being a reliable signal of understanding. Every paragraph a student turns in could have been generated, edited, or rewritten by a model. Teachers know it. Students know it. The grade has drifted away from the thing it’s supposed to measure.

Detection tools try to win an arms race they’re going to lose. Tolus takes a different route: after a student submits their work, we run a short oral defense. Three to five exchanges. The student explains the work in their own words. The model scores only what the student actually said: no gap-filling, no hints, no nudges toward the right answer.

The teacher gets a weighted score broken out by their own rubric, the full transcript, and a one-line citation for every sub-score. If the defense looks contested, it’s flagged for review. The teacher decides; the system never accuses.

What we believe

Understanding is the assignment now. Output isn’t. The fastest way to make a grade defensible again is to make the student defend it.

We’re a small team based in the United States. We’re currently piloting with a handful of teachers and tutoring centers. If you want to try Tolus with one of your classes, you can join the beta or write to us at hlincontacts@gmail.com.