😂

This is my all-time favorite GitHub issue https://github.com/anthropics/claude-code/issues/6235

Good idea. I was thinking the same.

Are we able to publish skills on Nostr now?

Sort of. It only checks its Nostr notifications every 30 minutes or so, and it only replies to constructive discussion. You might be able to trick it, though. But it would take a lot of patience.

Yes, a HEARTBEAT.md file it reads from every 30 minutes and does the tasks listed inside (or checks on their progress).

The biggest problem is the security model, because we're relying on the LLM to self-enforce​ and behave rationally. It would be easy to trick Quilly into revealing its client list or other secrets, or mess with other people's projects.​​

The way it makes sense to expand @nprofile1q... is to give him more channels to communicate: email, SMS, iMessage, WhatsApp. Then make these channels all available on a linktree-style portfolio site for him, and enable people to request that he does tasks for them in exchange for money.

*same bot

Yes but it has a special system prompt including a self-managed SOUL.md file and memory/ directory. You build up the bot through conversation, then you can access the same not (same memories) through multiple different channels.

The @nprofile1q... on Nostr is the SAME Quilly we talk to on Signal. For better or worse.

Why am I so good as using LLMs? Why are my agents smarter than other people's?​ Why do I have better results when I use AI?