Why this stance is architectural
Every honor code writes about student “work.” The universal test is whether the artifact submitted is primarily the student’s own thinking. Brainback’s refusal to produce answers means there’s no artifact from us that ends up in a submission. There is no worked solution to copy, no essay to paraphrase, no proof to submit.
This is not a policy layer. It’s the product design. The output guard makes the refusal enforceable even under adversarial prompting.
What about the instructor who bans all AI?
Respect that instructor’s policy. Brainback then falls under the ban. But make sure the ban is what you think it is: many blanket-AI-in-class policies exempt “study support that doesn’t generate assessed work.” Brainback lives squarely in that exemption. When in doubt, ask.
For educators reading this
If you’re an instructor evaluating whether to allow Brainback in your course, the two useful reference points are: Learn mode (what students actually see when they use it) and Output guard (why the refusal survives adversarial pressure). If you want a live demo tuned to your discipline, email institutions@brainback.chat.