Wave K — SoA research, 10 net-new features¶
Date: 2026 · Tier: design/ — public engineering research · Author: Mohsen Seyedkazemi Ardebili Companion: fed adr-v2-085 through adr-v2-094. See the waves index.
A snapshot of the field, kept as the research record behind the ADRs it fed. Feature status should be checked against
yazses featuresand the linked ADRs, not this note.
All on-device, off by default, distinct from the 82 features existing before this wave (v2 + Waves D-J). Ranked strongest-first; anchors web-verified at the time (2025-2026). Prioritized for accessibility and "needs local keyboard/window/selection/clipboard state" — the class of feature a cloud tool can't reach.
- Chorded Shortcut Synthesis — parse an arbitrary spoken modifier+key chord on the fly ("press control shift P", "escape twice", "hit F5") and inject it — not a pre-registered macro. Anchor: Apple Vocal Shortcuts (iOS 18, on-device), Talon chords, Apple Voice Control. Pure:
parse_chord(text)→ KeyChords, reuses the existing key-sequence injector. The biggest motor win in the set. → adr-v2-085. - Focus-Class Auto-Profile Switching — the daemon watches the focused window's app class and auto-activates the matching profile (grammar/vocab/cleanup/verbatim/injector). Anchor: Talon app-context, wlrctl/GNOME Window Calls/KWin/AT-SPI. Pure:
resolve_profile(window_class, map); focus-poller deferred to the platform layer. Distinct from manual per-app modes (automatic). → adr-v2-094. - Voice Undo/Redo Timeline — a ring of the daemon's own injection events so you can undo/redo its output across bursts by voice ("undo the last sentence", "go back before I said X"), even where Ctrl+Z is unreliable. Anchor: VoiceRev / Commanding-and-Re-Dictation TOCHI. Pure:
InjectionTimeline(append + undo(scope)/redo → backspace/retype deltas). The highest-leverage core in the set (also unlocks #8). → adr-v2-089. - Voice Case & Identifier Transform on Selection — "make this snake_case / Title Case / camelCase" over the clipboard selection. Anchor: Cursorless format actions, Serenade. Pure:
transform_case(text, style); uses the existing clipboard. Distinct from Spoken Code Mode (dictates new vs transforms existing). → adr-v2-087. - Auto-Pairing & Wrap-Selection for Code — balance
() [] {} "" '' ``, "wrap this in parens". Anchor: Serenade enclosure handling, Cursorless wrap. Pure:balance_delimiters+wrap`. Balancing/wrapping semantics literal dictation lacks. → adr-v2-088. - Inline Compute — "what's 15% of 240" → types 36; date math. Anchor: Apple Math Notes (iOS 18). Pure:
evaluate(expr)safe arithmetic/percent grammar (ast); sympy/date-math deferred. Distinct from unit conversion/LaTeX/temporal (evaluates vs converts/formats). → adr-v2-086. - Spoken Table → CSV / Field Data Entry — "row: Ada, 1815, London" → tab/comma cells + cadence (cell→Tab, row→Enter). Anchor: Vocal Forms (IJSREM 2025), VaaniSevak offline. Pure:
rows_to_delimited+ cadence machine. Distinct from Spreadsheet (nav vs bulk entry). → adr-v2-091. - Session Bookmarks & Resume — "bookmark here", "jump to my last bookmark"; anchors map to timeline positions (builds on #3). Anchor: VoiceRev / Just-Speak-It UIST. Pure:
BookmarkStore. Distinct from Spoken Recall (positions vs content). → adr-v2-090. - Word-Count & Writing-Goal Tracker — "how many words so far?", spoken goal progress via read-back. Anchor: eyes-free editing (UIST). Pure:
WordGoalTracker. Eyes-free utility. → adr-v2-092. - Local Voice Timer / Break Reminder — "remind me in 20 minutes to stretch", pomodoro; local notification. Anchor: Stretchly microbreaks, MTD/RSI vocal-rest literature. Pure:
parse_duration(text)+TimerScheduler; notification backend deferred. Distinct from Vocal-Strain Guard (user-set timer vs strain signal). → adr-v2-093.
Ship-now pure (do first)¶
1 parse_chord, #6 evaluate, #4 transform_case, #5 balance_delimiters/wrap — cleanest¶
dependency-free cores; #3 InjectionTimeline is the highest-leverage slightly-larger core (unlocks #8). Caveats from the original sweep: #2's auto-switch is anchored to Talon app-context (an ongoing product, not one paper); #8/#9 lean on established eyes-free-editing literature rather than a single result.
Citations here have not been re-verified against research/verify_refs.py.