https://docs.factory.ai/cli/getting-started/quickstart is very good. they just totally copied claude code but just made it better. much more context efficient. they also have a max plan and give you like 2,000,000 tokens free if you want to test. closed source tho.

Nice ideas

give me links

This is something i vibecoded a few weeks ago. just a static prototype and doesn't work yet, but i keep thinking about it and might actually implement Idea is that I would define a javascript DSL for coding agent state machines (e.g. "create 4 agents to all implement this task, then launch 2 reviewer, then have the reviewers argue, then choose one branch and make a pr"). maybe it would just be xstate (https://stately.ai/docs/xstate). User would explain workflow they want in natural languages and a state machine workflow would be generated automatically. they could inspect it -- should be short like 100 LOC or less. Could ask for changes in natural language. They see a nice mermaid style diagram. Then they could launch it and the state machine runtime would orchestrate this workflow. The state machine templates could be easily shared since they're just short text snippets

Yea I would just push replit as far as you can. If you hit a wall feel free to reach out again.

This sounds extremely useful

One really awesome thing about vibecoding is that you can do it for nonchalantly 15 minutes and actually be productive. With real coding I always needed to dial in for at least a few hours.

Also, have you ever looked at https://moq.dev/ ? I think it could be a really nice alternative to webrtc. It's far simpler and flexible. The best demo I made using it was https://github.com/justinmoon/av-demo. This does uses Marmot/Whitenoise encryption stack on audio calls. I think this would be a really nice stack for nostr a/v work. Happy to collaborate if you're interested!

I just vibe-coded a browser that can run react.js

Very cool. How does the encryption work?

Trick for when you want the AI to make a good spec

SICK!