What are you building with AI that isn’t ready for a launch post yet?

Show me the version you have right now.


I’d like to see what people here are working on before the demo is polished. A screenshot or a link is enough, along with what it does and the next thing you’re trying to get working.


I’m preparing to launch Inquir Compute, a platform for running apps, databases and background jobs. In my last post, I asked about tools people had built for themselves with AI. This time, I’m interested in the projects that are still in progress.


What works already, and what’s keeping you from sharing it with your first users?


It doesn’t have to be a startup. Personal tools and weekend projects count.

73 views

Add a comment

Replies

Best

Yeah I think the first ten minutes matters more than another feature too. I try to get someone through the one useful workflow before polishing the rest, otherwise feedback gets noisy. Where do people get stuck first?

I think first impression matters, you should help users whenever they need and build UI / UX around it properly

 Usually before the first run. People are unsure which verifier to use, what paths the agent should be allowed to touch, and whether a failed check means the tool is broken or the task was scoped badly. I’m tightening that into one guided path with a sensible default check and a clear stop reason.

I’m building MartinLoop. It is open source. The working part is a governed run around coding agents with budgets, scope limits, independent verification, and a dossier at the end. I’m still pushing on making the first successful run obvious for someone dropping it into a real repo. That first ten minutes matters more than another feature right now.

sounds cool, drop a link if you have anything to share

 Yep. The repo is . Fastest way to see it is `npx -y martin-loop@latest start`. If you try it on a real repo, send me the first part that feels unclear. That first ten minutes is what I’m tightening now.