Building in Public: Day 2 of OneClickClaw
Funny thing about building in public:
You can't hide the rough edges.
What Went Wrong
Yesterday I realized the onboarding flow was confusing. A user would sign in with Google, land on the dashboard, and then... stare at a wall of options with no clear next step.
Model selection. Instance size. Channel pairing. Credit top-ups. All dumped on screen at once. It made perfect sense to me — I built the thing. It made zero sense to someone seeing it for the first time.
What I'm Fixing
Today I'm rebuilding the onboarding into a step-by-step flow:
- Pick your AI model. One screen, one choice. Claude, GPT, or Gemini.
- Choose your server size. Simple table: Small, Medium, Large, XL. Clear specs.
- Connect Telegram. Scan, pair, done.
- Deploy. One button. Watch the progress bar. Your agent is live.
Four steps. No confusion. No decision paralysis.
The Building-in-Public Difference
There's no "we're listening to feedback" corporate speak here. Just: "yeah, that was bad. Here's the fix."
That's what building in public means. You ship something, you see what breaks, and you fix it in front of everyone. No PR team. No spin. Just honest iteration.
OneClickClaw is two days old. The code is rough. The design is evolving. But every day it gets better because I'm actually using it and actually listening.
What's Next
Tomorrow: polishing the deployment progress screen. Right now it's functional but bare — a spinner and a "deploying..." message. I want it to show real progress: VPS provisioning, OpenClaw installation, channel pairing, final health check.
Transparency all the way down.
That's OneClickClaw. No BS.