June 8, 2026
Build 120 - Privacy-Vertical Dictionaries
- 5 new industry dictionary packs New
- Therapy and Mental Health: 328 terms (DSM-5, CBT, DBT, EMDR, medications, billing codes)
- Government and Military: 206 terms (security clearances, ITAR, classified terminology)
- Accounting and Tax: 159 terms (tax forms, audit language, compliance)
- Human Resources: 151 terms (employment law, benefits, performance management)
- Insurance: 133 terms (claims, underwriting, policy language)
- Legal dictionary expanded from 236 to 294 entries Improved
- Medical dictionary expanded from 230 to 304 entries Improved
- Finance dictionary expanded from 158 to 234 entries Improved
- Total coverage: 2,800 terms across 13 professional packs
June 6, 2026
Build 118 - Voice Commands
- 19 voice commands for hands-free formatting New
- Say "new line" for line breaks, "new paragraph" for paragraph breaks
- Say "bullet point" to start formatted lists
- Say "open quote" / "close quote" for quotation marks
- Punctuation commands: colon, semicolon, exclamation point, ellipsis
- Symbol commands: hashtag, at sign, ampersand
- Auto-capitalizes after voice command line breaks
June 4, 2026
Build 114 - Lists and Web App Formatting
- Auto-numbered lists New
- Say "number one, do this. Number two, do that." and get a real numbered list
- Two-marker safety gate prevents false positives
- Line breaks in web apps - paragraphs paste correctly in Gmail, Notion, Slack Fixed
- Previously line breaks were collapsed to spaces in browser-based apps
June 2, 2026
Build 111 - Per-App Profiles
- Per-app profiles - PTT detects which app you're dictating into New
- App icon appears on the recording pill showing context
- Customize formatting settings per app (different rules for Gmail vs Slack vs Notes)
- Click-to-copy in menu bar - click anywhere on the last transcription preview to copy Improved
May 30, 2026
Build 110 - Stability and Speed
- Silence hallucination filter New
- Blocks 20+ known Whisper phantom phrases when you accidentally hit the hotkey without speaking
- Model preloads on launch - first recording is instant, no "loading model" wait Improved
- Reduced memory footprint during idle state