Loading
Brainback is designed to be compatible with your assignments, your honor code, and your students' privacy. Institutional pilots available every semester, from single-course trials to district-wide rollouts.
Every student using Brainback is producing their own reasoning, coached. The output is theirs, always. Four properties make it structurally compatible with the way you already teach — and with the way your honor code is written.
Anecdotal until we have proper controlled data (see the roadmap). But every pilot instructor we’ve talked to reports some subset of the pattern below.
| behavior | with brainback | answer machines |
|---|---|---|
| Homework completion rate | Higher — attempt-gate ensures kids engage before requesting help. | Higher via copy-paste; lower authentic engagement. |
| Exam performance | Higher on reasoning-heavy exams — students trained on the reasoning. | Lower on exams that don't reward pattern-recognition. |
| Office-hour productivity | Better — students arrive with specific stuck-points, not blank slates. | Same as pre-AI. |
| Cold-call quality (case-method) | Notably better — students actually engaged with the material. | Notably worse — AI-briefed cases don't survive interrogation. |
| Academic-integrity risk | Low — no worked answers to submit; no shared corpus. | High if students use ChatGPT/Chegg for graded work. |
| Student privacy | Content isolated from educator dashboards. Effort signal only. | Varies — Chegg posts questions publicly for tutors to answer. |
The evidence base for retrieval practice and desirable difficulty is a century old. Brainback operationalizes it in three moves — and each one is a lever educators have used for decades.
Ebbinghaus (1885) mapped the forgetting curve. Roediger & Karpicke (2006) showed retrieval > re-reading. Bjork & Bjork (2011) named the paradox: harder now, better later. Chi et al. (1994) on self-explanation; McDaniel & Little (2019) on interleaving. The literature is unanimous.
Brainback operationalizes that in three moves — attempt-gate before help, hint ladders that never give the answer, spaced retrieval built from the student’s own errors. Each move is a lever educators have used for decades; the software makes it default rather than heroic.
We’re rolling into pilots now. Once we have real numbers from real classrooms — with proper controls, sample sizes, and IRB oversight where required — we’ll publish them at /educators/research.
Your assignments. Your rubric. Your honor code. Your grade book. Your office hours. Your class discussions. Brainback is a study tool for students outside your classroom — it doesn’t enter the grading loop, doesn’t interface with your LMS unless you want it to, and doesn’t change what you teach. It changes what happens at 10pm the night before your homework is due.
Six steps from email to deployed cohort. Free for the first course, one semester.
District / campus rollouts. The published DPA, tiered pricing, a measurement plan.
Curated reading and open questions for instructors deciding to pilot.
The attempt gate is enforced in the API, not asked for in a prompt. A student cannot skip it, and neither can we.
→for educators
Email institutions@brainback.chat or head to /for-schools for pricing and a DPA. Deployment takes 2–4 weeks.