Sprint Brain

A model of how you write, kept on your device.

Sprint Brain is the part of SprintText that makes it yours. It is not a copy of your messages. It is a compact picture of your vocabulary, your recurring phrases, the corrections you keep reversing and which of your shortcuts actually earn their keep.

How it works

In, stored, out.

Three columns, and the middle one is the whole argument: what SprintText keeps is derived, not verbatim.

Signals in

  • A shortcut expanded, and how many keystrokes it saved
  • A shortcut undone straight after firing
  • A suggestion accepted or ignored
  • An autocorrect you reversed
  • Vocabulary you use, only with continuous learning switched on

The Brain

Counts, weights and a vocabulary — folded together on the device and updated as you type.

Not stored: the sentence around a word, which app you were in, or a transcript of what you typed. Raw typing retention is a separate switch that is off by default.

Behaviour out

  • Which shortcut the suggestion bar offers first
  • Which phrase Sprint Complete predicts next
  • Which corrections stop being made
  • The style My Voice writes in
Illustration — a diagram of the Sprint Brain data flow, not a screenshot of the app.

What it learns

Four kinds of knowledge.

  • Your vocabulary

    Names, product names, jargon, the surnames of everyone you work with. The words that a general dictionary keeps trying to correct into something else.

  • Your phrases

    The clauses you assemble again and again. Recognising them is what lets a keyboard offer you a sentence rather than a word.

  • Your corrections

    Every time you undo the same autocorrect is a vote. Enough votes and SprintText stops making it. This is the fastest-acting kind of learning, and the most satisfying.

  • Your shortcuts, ranked

    Usage counts and last-used dates decide which of your triggers gets a slot in the suggestion bar when several match what you have typed.

Sprint Trainer

Deep Train: skip the first two months.

Learning from live typing is slow at the start, because it has to wait for you. Deep Train reads writing you have already done, so the Brain starts out knowing your voice.

  1. Choose your samples

    Exported notes, old drafts, documents, a folder of things you have written. You decide what goes in, one batch at a time.

  2. Review what was found

    Sprint Trainer shows the vocabulary and phrases it extracted before anything is committed to the Brain.

  3. Keep or discard

    Accept the batch, or throw it away. A training run you regret can be removed without unpicking the rest of the Brain.

Control

Your Brain, your switches.

Everything about learning is opt-in, and everything that is learned can be looked at and removed.

  • Continuous learning is off by default. Until you turn it on, SprintText ranks the shortcuts you wrote yourself and learns nothing from your ordinary typing.
  • Raw typing retention is a separate switch, also off. The app explains why turning it on is rarely worth it rather than burying it in a settings screen.
  • Cloud AI is a third switch, also off. The Brain does its work on the device; the cloud is only involved when you invoke an AI action.
  • Sync is a fourth switch, also off. Nothing leaves your phone for another device unless you ask for that.
  • Reset means reset. Clearing the Brain removes the learned model. It does not quietly keep a shadow copy.

Questions

About the Brain.

Does Sprint Brain store what I type?

Not by default. What it keeps are derived signals — counts of events, weights, and a vocabulary of words you use. There is no field for the surrounding sentence or the app you were typing in. Retaining raw typed text is a separate setting that ships switched off.

Does the Brain go to a server?

Only if you switch on sync, and then only between your own devices, encrypted. Learning itself happens on the device. Cloud AI actions are separate again: they send the text needed for the action you invoked, not your Brain.

How long before it feels personalised?

Honestly: it depends on how much you type and whether you use Deep Train. Shortcuts you write yourself work from the first keystroke, which is why we suggest starting there. We would rather not quote a number we cannot stand behind.

Can I see what it has learned?

That is what the Sprint Brain dashboard is for — words and phrases learned, keystrokes saved, and which shortcuts are pulling their weight. Every figure is counted from something that actually happened; where a number is derived rather than counted, such as time saved, it says so.