Roadmap

haptics.md runs on two threads. The core thread is the format, the compiler, and the tools that make touch feedback portable and copy-ready. The accessibility thread is the north star: haptic braille and tactile guidance for blind and low-vision users. The core thread moves at engineering speed; the accessibility thread moves at the speed of co-design, and deliberately slower — it is the most careful work in the project.

Dates are targets, not promises. The durable core is the .hap.md format and the frozen intent vocabulary; everything else compiles from it. To influence the roadmap, open an issue or a discussion.

Legend: ▸ core / product · ◆ accessibility / mission · ○ tooling & distribution

2026 Q3 — v0 and launch

• ▸ Extend the data model (non-breaking) so every pattern emits a .hap.md; hand-author and device-verify a small seed set on iOS and real low-end Android. Ship the feel-it playground (Android-live), the “Copy as {platform}” switcher, the draft haptic/0.1-draft spec + JSON Schema + CI validator, the five agent skills, and the npx haptics.md CLI. Apache-2.0, governance, and contribution scaffolding in place. Deploy on Cloudflare Pages. Show HN + a 20-second real-phone-buzz video.
• ◆ Publish the braille draft as an open proposal and a call for collaboration — the braille/0.1-draft encoding, the developer guide, and the add-haptic-braille skill, all labeled provisional and explicitly not yet co-designed. This opens the door; it is not a feature.

2026 Q4 — v0.2, correctness as the product

• ▸ Extract @haptics-md/compiler (parse → IR → per-target emit) and a golden-fixture conformance suite; real AHAP + Android VibrationEffect.Composition emitters; publish the JSON Schema for editor + CI validation.
• ○ Tactons — ship a curated pack of learnable haptic notification signatures (message vs. call vs. alarm) as a registry category with docs. No new tech; pure curation and taste.

2027 Q1 — v0.3 (agents) and the braille track begins

• ▸ Wrap the compiler in @haptics-md/mcp + a hosted endpoint with .well-known/mcp.json. The agent layer goes live: telling an agent “add a success haptic” resolves through haptics.md.
• ◆ Braille, phase 1 — research and draft spec. Seat a standing accessibility advisory (majority blind/low-vision, braille-literate, multi-locale); open engagement with braille authorities (Unicode; BANA/UEB; Comisión Braille Española / ONCE and ULAC). Gate: no default is called “recommended” and nothing is labeled beyond “draft” until the advisory is seated and has reviewed it.

2027 Q2 — v1 surfaces, the studio, and the braille prototype

• ▸ Open the two-tier registry (curated core + community) and community adapters; React Native and Flutter as intent-tier targets, honestly labeled; staggered per-platform launches.
• ◆ Braille, phase 2 — native prototype. iOS Core Haptics temporal-cell renderer (sharpness-marked boundaries) + the VBraille spatial-via-touch mode + a generator for English Grade 1 letters and digits. Everything labeled “prototype”; the canonical-label invariant enforced in the schema before any file is published.
• ◆ Wayfinding — braille’s sibling: directional / turn-by-turn haptic cues, co-designed on the same cadence (a small codebook of single cells, not spelled words).
• ○ Haptics studio — the no-code, drag-and-drop visual editor whose output is a .hap.md (client-side; previews via the compiler). Studio output feeds the registry.
• ○ Smartwatch — Apple Watch (WKInterfaceDevice) and Wear OS as cue-tier surfaces. Front-loaded on purpose: the wrist is the best surface for the whole mission — always on skin, so braille, wayfinding, and tactons all land better there.

2027 Q3 — v1.x formats, and the first braille user testing

• ▸ Compile-to-formats as first-class targets: AHAP passthrough, Lofelt .haptic, MPEG-I HJIF; gamepad / DualSense rumble escape hatches; @haptics-md/react runtime.
• ◆ Braille, phase 3 — first blind-user co-design and testing round. The full arc (contextual inquiry, workshops, iterative prototypes) with braille-literate participants across congenital/late blindness, Grade 1/Grade 2, and a deliberate range of tactile acuity. Gate: results — including negative results and imperceptibility findings — are published before any release language is used. No release is cut this quarter; the deliverable is honest evidence.
• ○ Figma plugin — designers specify haptics in the design handoff and export .hap.md. Needs the compiler stable, which it now is.

2027 Q4 — hardening and honest widening

• ▸ Stability pass; a VS Code / Cursor extension (inline .hap.md preview + “feel on connected phone”); the “haptic of the week” content loop.
• ◆ Braille, phase 4 — retune and widen. Retune defaults from the testing findings; add the Android VibrationEffect emitter with hasAmplitudeControl() / areAllPrimitivesSupported() guards and encoding-not-just-amplitude degradation; add contracted Grade 2 and 8-dot handling by deferring translation to liblouis upstream; add the best-effort Android-Chrome web rung with an explicit no-op on iPhone Safari. Gate: the public claims register is live; every capability is paired with its limit; no web-parity language permitted.

2028 Q1 — first validated braille release (locale-gated)

• ◆ Ship English braille support only if the advisory and the English braille authority sign off, accompanied by published longitudinal usability data (task success, error, comfort, fatigue, perceptibility across acuity) and a wayfinding codebook. Gate: a locale ships on advisory + braille-authority sign-off, never on a marketing date. If the evidence isn’t there, it doesn’t ship.

2028 Q2+ — more locales, and stabilization

• ◆ Add locales one at a time, each co-designed with that locale’s braille authority and blind community (Spanish / Latin America via Comisión Braille Española / ONCE and ULAC as the candidate second locale); publish per-locale evidence. No auto-porting from English — every locale repeats research → prototype → testing → validated release.
• ▸ Target the WICG Web Haptics API as a compile target the day it ships. Move to vendor-neutral, multi-maintainer governance (see GOVERNANCE.md). Freeze haptic/1.0 — after which the metadata requirements, the canonical-label invariant, and the honesty rules carry full RFC-2119 weight.

How to influence this

Open an issue or a discussion. Accessibility and braille decisions are made with the advisory and the relevant braille authorities, not unilaterally — if you read braille or work in accessibility, that is exactly the input the mission thread needs.