Desktop
- ADE Code, without the flicker. Selecting a lane in
ade codeno longer reflows the whole drawer — lane cards are single-line, the expanded and compact chat lists render through one shared row, and the TUI reserves viewport rows to avoid Ink full-screen redraws. Provider glyphs and colors now match the model picker one-to-one, grid navigation escapes cleanly with Tab, and terminal previews dedupe their frames. - Multi-question approvals in the terminal. When an agent asks several questions at once,
ade codeshows a proper multi-question selector — arrow keys to choose an option and move between questions, number keys to pick and submit — replacing the old single-question quick-select. - Leaner stored transcripts. Bulky chat transcript payloads are compacted in storage while keeping durable replay and live output intact, fixing cases where a transcript could return missing content.
- macOS VM removed. ADE’s macOS VM runtime, its UI, and the related
ade vmCLI commands have been removed.
iOS
- AI-named lanes. Auto-creating a lane on mobile now names it from your prompt using the host’s small AI model — for both Chat and CLI — with a deterministic fallback if the model is offline or times out. Lane creation is never blocked by naming.
- New-chat composer parity. The new-chat composer now matches an in-session thread: a single permission dropdown when space is tight, the model pill, and the fast-mode toggle, all from one shared control row, with fast mode honored on create.
- Smoother chat scrolling. Scrolling through Work chats is faster — assistant previews are cached for visible rows and deferred until after the stream merges, with per-session terminal fingerprints.
- Mobile push removed. Push notifications have been removed; the app relies on its live sync connection, and PR actions from widgets now navigate in-app.