Last week I built an MCP tool that gives agents access to search for code snippets on nostr. The change in the output LLMs produce with it is MASSIVE; they no longer worked based on a large blob of unweighted code; now I can point them to see EXACTLY how I do a particular implementation under very precise conditions. Now whenever I see the LLM making a mistake I correct it, publish a code snippet, and the LLM never makes the mistake again. It’s absolutely wild to see LLMs using interfaces in NDK I literally just implemented and getting their uses right first time. This is not only useful for nostr development. This is useful for ALL development. And requires cryptographic identity, web-of-trust, censorship-resistance and money. All building blocks that are in our very robust toolbox. The future is very bright

Replies (40)

HoloKat's avatar
HoloKat 9 months ago
How to access? My bot always struggles with NDK. Occasionally works but usually have to fight it
dackdel's avatar
dackdel 9 months ago
This is too bright. I need my sunglasses 😎
Oh yeah, this mcp tool also allows each agent yo create and manage their own pubkey, they get a Nutsack wallet which they can later use to zap useful code snippets they found (I instructed mine to zap 1 sat when they are able to use one without having to fix anything and 10 sats when it fits a very narrow/niche use case) The agents can also post on nostr and if I or one of the pubkeys I tell them to obey, replies, they will heed the instructions and work based on that feedback. nevent1qqsrqz2c0xv87lyvxg2lyvyklt9w6yjdqe2ffhm6qc96zdl27488jngpr9mhxue69uhhq7tjv9kkjepwve5kzar2v9nzucm0d50asheq nevent1qqsrqz2c0xv87lyvxg2lyvyklt9w6yjdqe2ffhm6qc96zdl27488jngpr9mhxue69uhhq7tjv9kkjepwve5kzar2v9nzucm0d50asheq
I’m publishing a tutorial tomorrow. A lot of people are using this Nextjs thing so I built a package of ndk to optimize for it. The goal is single-shot production-ready apps.
Reading your note in my downtime while my Cline is vibe coding a Nostr app.
👀
PABLOF7z's avatar PABLOF7z
Last week I built an MCP tool that gives agents access to search for code snippets on nostr. The change in the output LLMs produce with it is MASSIVE; they no longer worked based on a large blob of unweighted code; now I can point them to see EXACTLY how I do a particular implementation under very precise conditions. Now whenever I see the LLM making a mistake I correct it, publish a code snippet, and the LLM never makes the mistake again. It’s absolutely wild to see LLMs using interfaces in NDK I literally just implemented and getting their uses right first time. This is not only useful for nostr development. This is useful for ALL development. And requires cryptographic identity, web-of-trust, censorship-resistance and money. All building blocks that are in our very robust toolbox. The future is very bright
View quoted note →
Have been doing something similar in fiat mining world. Building out agents and giving them access to internal required macros as tools. If you have heavy standardization in the way you produce business deliverables this plays really well with faster initial code generation. Soooo deflationary!
Lightning News's avatar
Lightning News 9 months ago
⚡⚡⚡
PABLOF7z's avatar PABLOF7z
Last week I built an MCP tool that gives agents access to search for code snippets on nostr. The change in the output LLMs produce with it is MASSIVE; they no longer worked based on a large blob of unweighted code; now I can point them to see EXACTLY how I do a particular implementation under very precise conditions. Now whenever I see the LLM making a mistake I correct it, publish a code snippet, and the LLM never makes the mistake again. It’s absolutely wild to see LLMs using interfaces in NDK I literally just implemented and getting their uses right first time. This is not only useful for nostr development. This is useful for ALL development. And requires cryptographic identity, web-of-trust, censorship-resistance and money. All building blocks that are in our very robust toolbox. The future is very bright
View quoted note →
the axiom's avatar
the axiom 9 months ago
nextjs is a server-side solution, nostr apps should be client-side
Testing zap receipts for this note… we made six attempts to⚡zap this note, at sebastian@lnd.sebastix.com, over a period of 13 minutes. Six of the zaps were successfully paid... Please check for 6 satoshis received. However, we did find that only five of the payments produced zap receipts. This is a problem because the user who zapped you would NOT see an active ⚡icon after zapping. They might think the zap failed, and therefore might not zap you again.... Also.... your average zap time was 6319ms (6.3 seconds). We consider this zap time slow... if possible, zaps should be confirmed in under two seconds. (If time is too slow, other Nostr users might think your zaps are broken, might not zap you again.) We recommend that you use a cloud-based Lightning node to get more reliable zaps.