The Learner Needs an Autopilot, Not Hints - Dev Log

What should a learning tool tell you while you work, if not a hint? Testing the learn flow live on antiderivatives, right after unifying the learning model across calibration, learning and verification, with no LLM judge left in the loop. I work one example on the canvas: x^2 differentiates to 2x, so the antiderivative of 2x is x^2 + C, since any constant disappears when you differentiate. The flow infers where I am, but it stopped telling me what to do next. So the idea that came out of it is a harness box: a running answer to "what should I focus on now", based on the model and the best next action, instead of live hints. Learn more: https://academy-k.com/?utm_source=you...