HomeWorkManifestoJournalContact
Case study02 / FRAME

Frame

A BJJ training journal built against the forgetting curve.

ServicesNative · On-device
RoleDesign + build
StatusAlpha
Year2026
Frame — the position graph built from logged sessions
Your game, built from logged sessions
The work
02 / FRAME

Frame is a BJJ training journal built on a hard truth: you forget most of what you’re taught within days. Log a session in seconds by voice, and Frame holds every technique, brings it back just before you’d forget it, and shows you — in plain data — that you’re actually improving. Built for the belts who feel stuck, so they have a reason to keep turning up.

The problem
Every rung of the ladder leaks — for the same reason.
White → blue~1 in 10Everything is new, everything hurts, and the first real sign of progress is months away.
Blue → purple50–70%The gains go subtle — angles, timing, awareness. It has a name: the blue belt blues.
Purple & beyondQuiet driftThey rarely churn like blues. They burn out and disappear without saying anything.

Different belts, different excuses, one mechanism: when progress goes invisible between classes, people drift.

Belt-attrition figures are community estimates, not studies — we treat them as the shape of the problem, not proof of it. ~6M practitioners worldwide (Gymdesk); median UK adult membership £70/mo (Matador, 2026).

1For the individual

You forget most of what you are taught within days, and you cannot see yourself improving. Frame holds the detail and shows you the proof.

2For the room

The cheapest growth in jiu-jitsu is the member you already have, kept one more year. A room that can see its own progress is harder to leave.

Built against the forgetting curve

A single session throws four or five details at you — the grip, the angle, the small correction that made the sweep finally work — and within a week almost all of them are gone. Frame is built around that decay rather than despite it.

Everything you log comes back to you on a schedule: a few days later, then a week, then a month, timed to land just before you’d lose it. There’s nothing to revise and no flashcards to sit down with — the app works out what’s about to fade and puts it in front of you. Retention is the product, and it’s the one thing a notebook has never done.

Frame — how well techniques are sticking, measured at 71%
Fig. 01 · Frame · how it’s stickingThe forgetting curve, measured on your own sessions: 71% recall, and the point — 2–4 weeks — where it drops below half.

The log you’ll actually keep

Everyone knows they should keep a training journal. Almost nobody does, because writing one up after an hour of getting choked is the last thing you want to do.

Frame removes the friction: talk for thirty seconds in the car — “started in closed guard, hit a scissor sweep to mount, got swept, tapped to an armbar” — and it transcribes on-device and turns it into a structured session you just confirm.

Then it reads your game back to you: the positions you live in, the subs you land, the ones that keep catching you, and hands you one or two specific things to hunt next session. A plan, not a feed.

Frame — a session being captured as structured data
You log this
Frame — the three holes in your game, ranked from those sessions
It finds this
Fig. 02 · Frame · log this, get thatThirty seconds of structured capture on the left. On the right, the same data read back a week later as a ranked shortlist — what catches you most, whether it is getting worse, and one drill for each. Nothing was typed twice.

The app wears your belt

Frame is built for the belt journey and it shows it — the whole app takes on the colour of your rank and changes with you as you’re promoted. It runs offline-first and keeps your training history yours: saved the instant you log it, private, exportable, never sold.

The goal underneath every decision is simple and hard: make the progress you can’t feel visible, so the belts who’d otherwise quit have a reason to come back for the next stripe — not more features, but more people still on the mat a year from now.

Frame — the same screen at white belt, accent derived from rank
Fig. 03 · Frame · the same screen, one rank downIdentical data, white belt. Every accent in the app is derived from the rank you hold — so the interface tells you where you are before you read a word of it.
Frame — every promotion with the mat hours behind it, and what the next belt costs at the current pace
Fig. 04 · Frame · the long viewEvery promotion with the hours behind it, and what the next one costs at the pace you are actually training — the progress you cannot feel, made countable.
Art direction

Made before the app existed.

Concept film · not product footage
The ledger · kept / cut
Built by choosing what to leave out.

+Kept

  • A thirty-second voice log, transcribed on-device
  • Recall timed to the forgetting curve
  • An identity that wears your belt

Cut

  • Typing up a session after training
  • A social feed to scroll
  • Your training history living on our servers
One we killed

We scoped Frame typing-first, with voice as a nice-to-have. Every test said the same thing: if logging costs more than thirty seconds, nobody does it — and a journal nobody keeps is worthless. So voice became the product and typing the fallback. The exact opposite of how we’d planned it.

Scope. What is public here is the athlete's app — the first surface of a longer system, and the only part we will show until it has earned the next. The work behind it, and who it is being built with, stays off the site on purpose.

StackSwift · SwiftUI · on-device transcription
Timeline2026
ConstraintOffline-first. Exportable. Never sold.