This is a great idea! What did you use to vibe code it?
https://github.com/johnbnevin/noobstr.me/tree/main
NOOBSTR.ME !!!
Aptly named LOL
Website has propagated to all regions that I have checked. Let's Encrypt! Certificate pending.
Noobstr is designed to be an onboarding link. What is NOSTR? I don't know. Click here and see.
Bugs that still exist:
- SSL pending
- Settings menu: buttons hang off right side
- Code comments are AI generated, some may not be relevant, some more may be needed to be added by human, and some (most) are commenting on code that I don't understand well enough to audit :D
Best Practices Improvements:
- Contributors much smarter than me that can tell me how I or AI has bungled the code, help with convention, sanitize, reduce vulnerabilities, optimize the code, etc
- Ask AI to further help me optimize and enhance security / remove vulnerabilities
- Add @hodlbod 's book Building Nostr, https://building-nostr.coracle.social/ , https://github.com/nostr-protocol/nips , and other 'nostr best practices' information / primary sources into ai and ask it to look at code and make suggestions
Enhancements planned:
- Check more relays when searching by npub, 3-5. nostr.band Iris, Coracle, Phoenix
- resolve all media links to just the base domain + …; 'video' 'image' , based on filetype
- scan for strings that are wrapped that are not yet resolved to 'npub' or similar, long links, and garbage text so that they can be ellipsed
- Ability to click on note to go see note on njump.me Also, a 'See whole conversation' button on each… linking to one of a set of random web clients that displays conversations well
- clicking 'just go' with a feed already there, and with the same settings, should only load new notes without disturbing the reading order. right now it's refreshing the whole page.
- add a few noobie / normie friendly relays to the dropdown. Although this is prone to hubris and makes maintenance more unwieldy.
Questions:
- Hosting considerations?
- What's an appropriate number of notes to fetch, and how much should i rate limit happy clickers on the load button? It also calls out to fetch usernames and avatars.
- What size should I limit displayed avatars to for general purpose?
- What sort of performance hit will loading usernames on slow connections have? Think 3rd world older phones (high signal onboarding).
- how should I monitor or accept feedback about performance? It's hosted on a 3rd party vps right now with other websites.
- Who to talk to about getting this listed at places like https://nostrapps.com/ or grownostr.org at the appropriate time?
#vibecoding
#grownostr
nostr:npub18ams6ewn5aj2n3wt2qawzglx9mr4nzksxhvrdc4gzrecw7n5tvjqctp424 nostr:npub1nl8r463jkdtr0qu0k3dht03jt9t59cttk0j8gtxg9wea2russlnq2zf9d0 nostr:npub1aeh2zw4elewy5682lxc6xnlqzjnxksq303gwu2npfaxd49vmde6qcq4nwx
Discussion
Claude Sonnet 4. Simple questions I ask DuckDuckGo's AI so I don't have to start a whole new thing or count against my allowance.
I started off having it rewrite the code for me too much, anything else seemed way too overwhelming. But after using this prompt:
"Don't do it for me. Show me how to implement the change, one step at a time. Go slow, I'm dumb. Show me exactly what my code looks like now, and what exactly I should change. Explain everything. Ask me about options. Don't introduce security vulnerabilities."
I'm actually learning some things about being a developer. Well, a wannabe developer haha
I also just learned to slow down, take it one step at a time myself, and read it through, just to see what I can absorb by osmosis, even if almost nothing. And save each working copy as a different file so I always have a previous version to go back to that works.
Granted, I'm just building stuff that is one file, self contained so far. But the AI allows me to create something that works, whereas I would not otherwise have been able to. Thanks to you all for pioneering and paving the way!
I love this learning approach. Everyone prompting has such good ideas on how to do it better and learn something.
Instead of saving each file, which is obviously a natural tendency, it helps to learn how to use GitHub. I installed the desktop app and make sure to commit my changes often, and work in branches so I can experiment with new features without breaking anything I’ve already built. It’s been a game-changer.
💜💜💜
Does the Claude code integration help with this? Maybe nostr:npub1xr8tvnnnr9aqt9vv30vj4vreeq2mk38mlwe7khvhvmzjqlcghh6sr85uum knows
Yes, absolutely. That’s what I do now. Just make sure you are the one doing the commits yourself. If you ask Claude to do it for you, it will give itself credit in the commit. I learned that the hard way.
Can’t let the robot win