# YazSes v2.5.0-dev.1 — Wave I 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 I** — a fresh 2026 SoA round (`design/vision/v2-research/11-wave-i.md`) of 10
features distinct from the prior 66. This first drop ships the two cleanest pure cores:

## New features (all off by default)

- **Terminal Command Safety Gate** (ADR-v2-065, `[cmdsafety]`) — when dictation targets a
  terminal, a destructive command (`rm -rf`, `dd of=…`, `mkfs`, `git push --force`, `curl … | sh`,
  a fork bomb) is **held until you say "confirm"**, so a misrecognition can't fire an irreversible
  action. A safety guarantee no cloud dictation tool can offer — it has no view of your terminal.
- **Spoken Regex Builder** (ADR-v2-066, `[spokenregex]`) — build search patterns by voice:
  "four digits dash two digits" → `\d{4}-\d{2}`, "lines starting with a capital letter" → `^[A-Z]`,
  "one or more digits" → `\d+`. Feeds find dialogs and `grep`. A win for developers who can't
  type fluently.

`yazses features` now lists **77 capabilities**.

## Wave I backlog (heavier / hardware-gated)
Hard Contextual Biasing (trie hotword boosting), Diarized Conversation Capture + Rename-by-Voice,
Adaptive Latency Governor (speculative decoding), Per-Language Auto Model Switching, Document-Wide
Find-and-Replace, Structured-Markup Dictation, Voice Window/Workspace Management, Citation-by-Voice.

## Quality
1212 tests green across Linux × macOS × Windows on Python 3.11 and 3.12. Both Wave I cores are
unit-tested to 100% with **no new dependency**; the default install and the v1 dictation path are
unchanged.