v0.1 was a place you queried. v0.2 makes your brain follow you across chat surfaces, run your code while you sleep, and turn its own output into shareable artifacts.
Claude CLI, Claude Web, ChatGPT, and Telegram on your phone — same memory, same tools, same draft-then-confirm safety. Your brain speaks the open connector standard now, so any client that follows the spec inherits the same operating manual.
If you'd ask the agent to do it twice, write it once and let brains run it. An automation is a small program that runs on a schedule, against your brain, with the same tools the agent uses. Reads any source, writes drafts, calls its own LLM, keeps state — sandboxed, logged, killable.
Boards are spreadsheet-shaped datasets attached to a brain — every row JSON, every row queryable by the agent. Mini-sites are static, sandboxed HTML pages the agent generates from those boards. Both first-class. Both collaborative. Both shareable.
Nightly thinker & learning profile, tasks polish, BrainChat, Telegram bot, routing & ingestion, onboarding, web app, admin tooling — the wider surface that shipped alongside the three big moves.
/tasks page
Auto-extract from email · cal · chat
Auto-close on follow-up
Calendar invites no longer become reminders
BrainChat · live on stage
Auto-friend on login
"Ping anyone" defaults to BrainChat
Attachments on drafts
Telegram · claim-code flow
Daily token budget
Smarter calendar (drops passive subs)
Per-exchange capture
HTML formatting
list_calendar_events tool
Per-page incremental indexing
Routing v2 · access + audience-intent
Alias-aware identity
One-shot /install
Brains tour
Versioned CLAUDE.md
Devices view in admin
App shell · design tokens
Single home page
Mini-sites open full-screen
Tabbed admin dashboard
Per-call LLM telemetry
Services health probe
+ more