In beta · iOS
An iPhone keyboard that learns how you type and fixes your typos on the device.
For iPhone users whose keyboard keeps fixing the wrong words: names, slang, or what they actually meant.
Your words. Fewer typos.
A keyboard can be fast and still be frustrating when it replaces names, slang or intentional wording. A system that only fixes one word at a time can also miss an earlier typo whose meaning becomes clear later in the sentence. The challenge is to use later context without freely rewriting what the person said.
Idio is an iOS keyboard for recovering the words a person intended to type while keeping the person in control. Its core work is local: touch evidence, language context, bounded correction and reversible edits.
01Screen · Phone
Home: whether the keyboard is ready, and a place to try it.
iOS simulator, seeded test data, 21 September 2026.
02Screen · Phone
Every correction can be taught, kept, refused or forgotten.
iOS simulator, seeded test data, 21 September 2026.
03Screen · Phone
Teach Hands: where your thumbs actually land, key by key.
iOS simulator. The tap data is seeded by a test, not a person's taps.
04Screen · Phone
What changed because of you, and whether it is in use yet.
iOS simulator, seeded test data, 21 September 2026.
05Screen · Phone
Appearance changes the board, never what it learns.
iOS simulator, 21 September 2026.
How it is used
01 · System
The correction path runs on the phone. The model is optional and can be late.
Select a part to see what it is responsible for and what it connects to.
Tap positions and the text being typed, inside the keyboard extension.
→ Personal motor model: calibration taps
→ Intent decoder: typed span
Teach Hands evidence per hand position, speed and layout.
Trade-off: Coverage is not accuracy, so more samples are never shown as better typing.
→ Intent decoder: personal evidence
Bounded search over candidate readings using language, personal, motor and context evidence.
Trade-off: A safety boundary can leave a typo unfixed rather than guess.
→ Ownership and protection: best reading
Protected words, kept originals and owned-document checks decide what may change.
→ Reversible edit: allowed change
Apple Foundation Models for optional sentence review, when available.
Trade-off: A late answer is rejected if the text changed underneath it.
→ Ownership and protection: proposed review
The change lands with Undo, Keep Original and Never Change one tap away.
→ Personal motor model: trustworthy labels only
Idio uses native Swift and UIKit for the keyboard and companion app. The local intent decoder combines candidate, language, personal, motor and contextual evidence in a bounded search. Optional review can use Apple's on-device Foundation Models for constrained sentence review, alongside a narrower automatic cleanup path. Model availability is conditional, and deterministic behavior remains the fallback. The keyboard has no cloud dependency for its core correction path; account and sync foundations exist, but they are not released cloud features.
02
Teach Hands separates hand mode, speed round and keyboard geometry. Compatible evidence is reused rather than pooled into a misleading single universal accuracy score.
Candidate scores distinguish research, personal and population context. A bounded beam search produces possible readings; an unchanged reading remains meaningful.
Protected words, retained originals, owned-document checks and Undo are part of the product contract. A model preference is not permission to bypass those boundaries.
It offers explicit sentence review and a narrower automatic cleanup after a pause. Availability, timeouts and fresh-document checks matter; this is not an unrestricted cloud writing assistant.
03
A native keyboard extension and its companion app.
Swift · UIKit · iOS keyboard extension
Correction that runs on the device, with an optional on-device model.
Local intent decoding · Personal motor evidence · Apple Foundation Models
Sharing state with the extension, and purchases.
App Groups · StoreKit
04
Approach
The engine ranks bounded candidate readings and distinguishes personal touch evidence from broader language priors. Protected words and document ownership constrain what can change.
Trade-off
A safety boundary can leave a typo unfixed. That is preferable to pretending the system can always know the intended meaning; the user must retain an easy way to correct or undo it.
Approach
Teach Hands measures coverage within the relevant hand, speed and geometry channel. The current free-flow calibration regime rejects ambiguous or edited words as strong motor ground truth.
Trade-off
Coverage is not accuracy. A larger sample count does not prove better typing, and old calibration evidence cannot silently masquerade as evidence collected under a new labeling regime.
05
Links checked 21 September 2026
Recorded behavior Replayed from the test suite · nothing runs here
Switch between recorded cases from Idio's Swift tests: what was typed, what the keyboard is expected to produce, and what each correction control keeps.
Pick a sentence
Typed
teh
Idio's expected text
the
teh → the
Settled at its own space; no later context needed.
iOS/Tests/RollingIntentTests.swift · testAnOrdinaryRepairNeverClaimsLaterContext · lines 170-180
Then you decide
Idio corrected “senes” to “sense”. Choose what happens next:
make sense n
Answers quote the source notes and cite them.
