Kunal Abichandani

What if we could manipulate UI at the speed of thought?

Built this voice coffee ordering system with @typesafeai Jev. The drink and any updates show up while I’m still mid-sentence, ~200 ms from word to screen.

Feels pretty magical!

I gave my agents a way to create Live Activities on my iPhone lock screen.

Their notifications were becoming noise, so I built Agent Activity to surface what matters, updating in real time

Some of my fav use cases with @chatgpt, @muse, codex:
- Birthdays of people I’m close with (from gcal + fb + linkedin), first thing in the morning
- What each agent is working on and what’s done
- Closest citibike dock with ebikes + train times before I head home
- Critical updates like root cause found for a bug, with a button to jump straight to the logs

Install the iOS app + connect the MCP in your agent(s). Your agents compose the layout and update a shared live activity as they work, or through recurring scheduled tasks.

Comment/DM if you want to try on testflight!

@Muse having its own app (vs others using iMessage) reminds me of Cursor building their own IDE instead of becoming a VS Code extension

Small things that make the UX feel complete: seeing the browser live, approvals, agent status, feed / ideas. IMO this will compound over time

Grateful to be part of the inaugural @a16z Design Engineer fellows.

Amazing crew 🙌

I built an agent that makes me justify my distractions.

- Texts me when I open distracting apps
- Gets firmer the more I use them, personalized to my motivations
- Blocks apps when I hit the limit
- Unblock requests only get approved if I can explain how they align with my goals