The gate is the first thing about Brainback that feels different, and it is the one rule with no exceptions: nothing happens until you have written an attempt of at least twenty characters. It is enforced in the route, not requested in the prompt, which means neither a clever message nor a future version of us can talk it away.
Twenty characters is a judgement, not a finding. It is long enough to exclude "idk", "help" and a pasted question mark, and short enough that a student who has genuinely started — one line of algebra, one sentence of a thesis — is never blocked. We would rather let a thin attempt through than refuse a real one.
An earlier version of this page claimed we had tested that. It described a silent experiment on a subset of beta users, reported effect sizes to the percentage point, and concluded that the gate was working. None of it happened: there was no beta, no experiment, and no data. It has been removed, and this paragraph is here instead because a research page that invents its own findings is worse than one that admits it has none.
What we would have to run, when there are enough students to run it honestly: a pre-registered comparison of attempt thresholds, measured on retention at one week rather than on messages per session — engagement is exactly the metric that would reward the wrong answer here. Any result we publish will arrive with its method, its sample size and its disclosure to the people in it.
Until then the argument for the gate is the published literature on retrieval practice and generation, which is strong and is not ours, plus the reasoning above. That is all we can claim, so it is all we will.
Written by Brainback Research. Published Oct 13, 2026. Filed under Method.