Roadmap

Built, building, next, and never

Including the last one, because a rejected feature is only worth anything if the reason stays written down where someone can check it.

Working now

In the firmware today

Every screen on the home page is one of these. Logic is covered by an automated test suite that runs on a desktop machine; hardware bring-up is still in progress.

WCA averaging

Ao5, Ao12, Mo3, rolling Ao50 and Ao100, session mean — with +2 and DNF handled the way the rulebook handles them.

The session

A rolling window of solves that survives a power cycle, with penalties, discard and undo on the newest solve.

Trend and spread

Two charts no other timer draws: are you getting faster, and are you consistent.

History browser

Every solve in the session, scrollable, with its penalty state.

The coach

Around fifty hand-written lines that appear when there is something worth saying.

Screen care

Dims after a long idle and wakes on a timer packet — because scrambling by hand is not inactivity.

On the bench right now

Cube type selection

The one we are actually building at the moment. Not in the firmware yet.

Being built

One device, a record for every cube in the bag

Pick the cube you are on, and everything is kept separately for it — the session, the history, the averages and the personal bests. A 3×3 evening and a Megaminx evening stop averaging into each other, and your 4×4 times stop dragging your 3×3 mean around.

Nobody who cubes owns one cube. Right now every solve lands in the same pile, which makes the pile close to meaningless the moment you pick up something different. This fixes that.

  • Choose the cube type, and the session follows it
  • Separate history per type
  • Separate Ao5, Ao12, Mo3 and rolling averages per type
  • Separate personal bests per type

Next

In the order we intend to build them

Ordered by value per unit of work, not by ambition. None of these is in the firmware today.

1 · Inspection

WCA 15-second inspection with the judge's calls at 8 and 12 seconds, +2 past 15 and DNF past 17. The biggest single gap, and the rules are exact, so there is a right answer to implement.

2 · Sound

Audible inspection calls and a confirmation when a solve stops. Inspection needs it anyway.

3 · A record book

Best Ao5, Ao12, Ao50 and Ao100 — not just the best single. The numbers a cuber actually quotes.

4 · Live averages

The rolling Ao5 and Ao12 while the round runs, next to your best ever, with the delta.

5 · Pace mode

Set a goal — sub-12 Ao5 — and the device tells you what you need on the remaining solves to still make it. The WCA trim makes that a real calculation you cannot do in your head mid-round. The one thing on this list nothing else does.

6 · Streaks and goals

"Seven sub-12 solves in a row." Trivial to compute, disproportionately motivating.

7 · Offline export

A QR code of your session on the screen; point a phone at it and you have the data. No radio, no cable, no account. Worth prototyping before we promise it.

8 · Custom profiles

Same cube, different discipline. A 3×3 solved one-handed is not a 3×3 solve, and a blindfolded one is different again — different times, different personal bests, different practice. Profiles keep a separate set of records for each way you solve, on top of the cube type itself.

9 · Scrambles, if they can be done properly

Last on the list, and by some distance the hardest. We are not ruling scrambles out — we are ruling out doing them badly. A random-move scramble is the easy version and is not what the WCA uses, so it will not ship. Random-state means a real solver with pruning tables, which is serious work on hardware this size. If it lands, the cross and F2L trainers that depend on it become possible too.

Not on the roadmap

Things we are not going to build

Each of these is a feature someone will suggest, and each has a specific reason it is a trap rather than a gap.

Times of day on your solves

There is no clock and no network. A device that stamps a solve with a time it invented is lying in the one place a record book must not. Solves are numbered instead.

Split times

The timer reports one time per solve. Cross/F2L/OLL/PLL splits need a second sensor or a press mid-solve, and you are not pressing anything mid-solve.

A conversational coach

The messages a cuber wants are all deterministic functions of numbers the device already has. There is nothing a language model would add except latency and doubt.

A millisecond result

The running number keeps its milliseconds because they are motion. A result truncates to the hundredth, because that is the time a delegate would accept. It is not a setting.

What would you build?

That list moves when cubers tell us it should. The waitlist form asks what you would build first — and puts you in line for one of the first production units while you are at it.

Join the waitlist