Robert Bjork's decades-long research on learning identified a paradox: the interventions that make immediate performance worse often make long-term retention better. Spacing your practice sessions makes each session feel harder. Interleaving problems makes each problem harder to solve. Testing before you're ready makes you feel unprepared. All three improve durable memory.
This is not a paradox once you understand the mechanism. Learning-in-the-moment (encoding) is not the same as learning-that-lasts (consolidation and retrieval strength). Difficulties that impair encoding often facilitate retrieval, because retrieval is what strengthens the memory trace.
Brainback's design leans hard into this principle. The attempt-gate is a desirable difficulty. The hint-ladder is calibrated to hold you at the edge of your competence without collapsing it. The spaced-recall queue interleaves subjects and spaces intervals so that every review is slightly harder than passive re-reading.
Users often describe the experience as 'the AI is making me work harder than ChatGPT does.' That is the point. That is exactly the point.
Written by Brainback Research. Published Feb 4, 2026. Filed under Learning science.