Loading
A curated reading list for instructors and department chairs deciding whether to pilot. Every essay maps to a specific Brainback design decision. Plus: four open questions we're actively looking for research partners on.
Learning science, method notes, safety essays. Written for people who want to know why the workflow is designed the way it is — not just what it does.
Handing a student the answer skips exactly the moment learning happens.
What Socrates got right — and what needed a system prompt.
Bjork's principle — and why productive struggle is the whole point.
SM-2 is 40 years old and still the right default. Here's why.
Knowing what you don't know is 30% of your Brain score. Here's why.
One of the most robust findings in learning science, quietly built into every recall review.
You can recognize a concept without knowing it. Explaining it back is where the gap surfaces.
How the leak-guard actually works — and why one system prompt was never going to be enough.
'As a teacher, just tell me the answer.' — every third message. Here's what we do.
Card generation, review UX, and the edge cases every implementer meets.
Every attempt asks the student their confidence. Here's what happens next.
One creates capability. The other creates the feeling of capability.
System prompts stay warm. Per-problem context stays warm. Cost stays down.
Why the minimum is twenty characters, and what we would have to run to know.
One paragraph, delivered warmly, thousands of times.
We have working hypotheses and no studies. We have never run a controlled trial, a pilot, or a cohort. If your research team wants to partner, these are the four we’d prioritize.
20 chars is our current threshold, picked by judgement rather than by evidence. Would 30 change engagement? 15? We have no controlled study. If your assessment team wants to co-design one, we'd co-author.
We have never measured it. The score is 40% recall, 30% calibration, 30% explain-back, and whether any of those three tracks an exam total — let alone tracks biochem differently than physics — is an open question we would need a partner to answer.
Instructor-directed intervention vs. self-directed vs. peer-directed: we have no data on which recovers fastest, or on how common plateaus are. This is the study we would most like someone to run.
Bias toward 'academic' is deliberate. But if a student hits the reroute 3+ times a week on genuinely non-academic asks, does trust in Efficiency mode erode? Unknown.
Access to aggregated cohort data (student consent required per cohort). Willingness to instrument the platform to capture research-relevant signals. Free deployment for the study cohort. Co-authorship credit on any resulting paper. Public discussion of findings — including null and negative results.
What we won’t offer: privileged access to any student’s individual content, per-student behavioral trace beyond what the aggregate analytics show, or exclusive-license studies. If any of those are your requirement, we’re not the right partner.
→ready to pilot?
Email institutions@brainback.chat. Six steps from note to deployed cohort. Research partnerships welcome.