# YazSes v2.6.0-dev.1 — Wave J opens (developer preview)

**Pre-release · not published** (publish workflows are pinned to `v0.x`/`v1.x`; v2 dev tags do
not ship). **v1.4.1 remains the stable release.** Every feature is **OFF by default**.

Opens **Wave J** — a fresh 2026 SoA round (`design/vision/v2-research/12-wave-j.md`) of 10 features
distinct from the prior 72. Two independent research streams converged on the flagship
(Confidence-Gated Re-Ask). This first drop ships the two cleanest fully-pure cores:

## New features (all off by default)

- **Phonetic Spelling Mode** (ADR-v2-075, `[spelling]`) — a character-exact mode for
  passwords/codes/IDs/filenames: "capital alpha bravo double lima" → `Abll`, with digits and
  symbols. The one thing free-form dictation always mangles. A large motor/low-vision win.
- **Voice Git Choreographer** (ADR-v2-076, `[gitvoice]`) — drive git by voice via a **structured
  argv grammar** (never free-form shell, so no injection risk); reversible ops (commit, checkout,
  merge) flow immediately, destructive ones (`push --force`, `reset --hard`, `branch -D`) wait for
  a spoken confirm, and the exact **undo command is always spoken**. Distinct from the Terminal
  Command Safety Gate — this *builds* git intent rather than classifying a formed command. No
  deferred backend: it's a plain `git` subprocess.

`yazses features` now lists **87 capabilities**.

## Wave J backlog
Confidence-Gated Re-Ask (flagship), Verbatim⇄Autoformat Live Toggle, Crowd-Proof Dictation
(target-speaker extraction), Self-Learning Correction Dictionary, Voice Fuzzy File Open, Voice
Jump-to-Symbol, Recording Import (batch transcription), Spoken Shell Pipeline Builder.

## Quality
1262 tests green across Linux × macOS × Windows on Python 3.11 and 3.12. Both Wave J cores are
unit-tested to 100% with **no new dependency**; the default install and the v1 dictation path are
unchanged.