Opened the waitlist for StareBrain — the AI phone agent I posted about yesterday
Quick update on StareBrain (the natural-language Android agent from
yesterday's post) — opened up a waitlist today.
To be clear about where things actually are: no working backend yet,
no LLM integration built. This isn't "come try the product." It's
"I want to build this with real people's input from day one, not
in isolation for 12 weeks and then find out what I got wrong."
Signing up gets you:
- Early access when there's something to click
- A say in which 5 commands ship first (SMS, calendar, device
toggles, app launch, calls are the current plan — open to being
wrong about that) - My honest build updates, including the boring/broken parts
If you're curious: https://starebrain.vercel.app/waitlist
Still chewing on the confirmation-flow question from yesterday's
thread — once/plan vs. once/action. Haven't decided. If you have
a strong opinion, it'll actually shape what gets built.
Love the transparency on where things actually are - "no working backend yet" is the kind of clarity most founders hide behind marketing copy. The 5-command prioritization by early users is the key insight though: you're not measuring success by launch day buzz, you're measuring by which signals matter to actual users. That constraint of "what do the first 5 need most" forces you to see which commands drive real adoption vs which ones just feel like they should exist.
appreciate that, and yeah I'd rather be upfront about where it actually is than oversell it and have people find out the hard way
the 5-command thing has been a genuinely useful constraint honestly. easy to list 15 things an agent "should" do, way harder to defend which 5 actually matter when someone's giving you real feedback instead of just imagining the product. a couple of the ones I assumed would be top picks (calendar stuff) haven't come up nearly as much as things like send-a-text and device toggles, which I didn't expect going in
still early enough that this could shift again once more people are actually testing it, but so far the simplest, most immediate commands are winning over the "impressive" ones