Avatar
Rod
1bda7e1f7396bda2d1ef99033da8fd2dc362810790df9be62f591038bb97c4d9
Startup founder and listco CEO, New Zealander (kiwi) expat living in Australia, #nostr #austrich, dad. - Nostr Blog https://rodbishop.npub.pro/ - Nostr FOSS https://github.com/r0d8lsh0p/ - Nostr Live Streaming nostr:npub1sh0spghk4yvy2d2v35kelw45qq4msk6zykaw4ds047e9slzs8r4qr7q2xa - Nostr AI nostr:npub1ahjpx53ewavp23g5zj9jgyfrpr8djmgjzg5mpe4xd0z69dqvq0kq2lf353 - Jayride (ASX:JAY) - Fishburners - #Bitcoin miner

Tested and working on kind 7 (like) and kind 5910 (DVM request)! Thanks for the awesome fast response.

Will put to work soon.

Personally I was looking forward to a personal spacecraft and a moon base but mate, you do you.

Demo Time:

https://v.nostr.build/VWYvyenZa62dkGBi.mp4

Claude Desktop communicates with a local Nostr Model Context Protocol Server.

The MCP server looks for Nip89s of MCP enabled data vending machines and provides them as tools to the LLM. When Claude finds it adequate it calls a tool (asks the user if it's ok to use the tool) and the MCP server makes a NIP90 request to the DVM and gives back the answer to the LLM. The LLM builds us a nice report based on the real time data provided by a DVM living on Nostr (in the case I run it on the left screen).

nostr:nevent1qqsx07t2rwzvae6tumn5j6lvugp79psyz27z9k53lw4h2jekszrgejqpz9mhxue69uhkummnw3ezuamfdejj7q3qr0d8u8mnj6769500nypnm28a9hpk9qg8jr0ehe30tygr3wuhcnvsxpqqqqqqz4q0990

Incredible stuff!

Claude Desktop (and Goose, and Cline, etc) can now use MCP tools made available over the Nostr network.

This is using the brand new DVM-MCP bridge by npub1nxa4tywfz9nqp7z9zp7nr7d4nchhclsf58lcqt5y782rmf2hefjquaa6q8

See his Github– https://github.com/believethehype/nostrdvm/blob/main/tests/mcp/nostr_dvmcp_server.js

Time to start deploying tools.

Replying to Avatar Lyn Alden

When it comes to AI, philosophical people often ask "What will happen to people if they lack work? Will they find it hard to find meaning in such a world of abundance?"

But there is a darker side to the question, which people intuit more than they say aloud.

In all prior technological history, new technologies changed the nature of human work but did not displace the need for human work. The fearful rightly ask: what happens if we make robots, utterly servile, that can outperform the majority of humans at most tasks with lower costs? Suppose they displace 70% or 80% of human labor to such an extent that 70% or 80% of humans cannot find another type of economic work relative to those bots.

Now, the way I see it, it's a lot harder to replace humans than most expect. Datacenter AI is not the same as mobile AI; it takes a couple more decades of Moore's law to put a datacenter supercomputer into a low-energy local robot, or it would otherwise rely on a sketchy and limited-bandwidth connection to a datacenter. And it takes extensive physical design and programming which is harder than VC bros tend to suppose. And humans are self-repairing for the most part, which is a rather fantastic trait for a robot. A human cell outcompetes all current human technology in terms of complexity. People massively over-index what robots are capable of within a given timeframe, in my view. We're nowhere near human-level robots for all tasks, even as we're close to them for some tasks.

But, the concept is close enough to be on our radar. We can envision it in a lifetime rather than in fantasy or far-off science fiction.

So back to my prior point, the darker side of the question is to ask how humans will treat other humans if they don't need them for anything. All of our empathetic instincts were developed in a world where we needed each other; needed our tribe. And the difference between the 20% most capable and 20% least capable in a tribe wasn't that huge.

But imagine our technology makes the bottom 20% economic contributes irrelevant. And then the next 20%. And then the next 20%, slowly moving up the spectrum.

What people fear, often subconsciously rather than being able to articulate the full idea, is that humanity will reach a point where robots can replace many people in any economic sense; they can do nothing that economicall outcomes a bot and earns an income other than through charity.

And specifically, they wonder what happens at the phase when this happens regarding those who own capital vs those that rely on their labor within their lifetimes. Scarce capital remains valuable for a period of time, so long as it can be held legally or otherwise, while labor becomes demonetized within that period. And as time progresses, weak holders of capital who spend more than they consume, also diminish due to lack of labor, and many imperfect forms of capital diminish. It might even be the case that those who own the robots are themselves insufficient, but at least they might own the codes that control them.

Thus, people ultimately fear extinction, or being collected into non-economic open-air prisons and given diminishing scraps, resulting in a slow extinction. And they fear it not from the robots themselves, but from the minority of humans who wield the robots.

There is no end to human wants, and so there is no end to human labour.

Average people today enjoy luxuries reserved for elites in times gone by.

In the future when average people enjoy the luxuries of today's elites, what will happen? Will they all be satisfied and living leisurely lives?

No. They will come up with something new to want.

Hi! Yes I do. And my apologies for the cheeky message, I just didn't want to trouble you with a DM.

Here it is, my first PR on a FOSS project. Perhaps I didn't submit it properly? If something is incorrectly done please let me know, it's guaranteed to be ignorance at my end.

https://github.com/bitvora/sw2/pull/1

You'll have to do better than that

Welcome to Nostr nostr:nprofile1qqsx7ke38jurn3ukuvjz7ercez6wacgj844cr42ek4qcu7854ql4saql73re9

Needed more DVM tool servers to test agents against.

Dad Jokes should reply to kind 1 notes instantly, and should have an annoying 10 minute or so delay in responding to DVM requests.

Need to build workflows to accommodate slow or asynchronous responses.

I think in your current UI setup you request the user input the npub, and that in the backend is used to specify the filter "#p":["hex pubkey"] ...?

I think that works great.

I would only need another field, similar to the current "Write" node layout, where you request the user input the Kind, with default to 1.

I use that a lot to write other kinds like 7 (reactions), 5910 (DVM request), 9734 (zap request), etc.

Your current mentions method will be good I think, the same method works for all kinds that I have seen; the mentioned user is tagged with "p" tag.

Kinds within the 5000-7000 range are defined with NIP-90. Within NIP-90 "p" tag defined as:

"p: Service Providers the customer is interested in. Other SPs MIGHT still choose to process the job"

That's the kind of mention I'm looking to work with.

https://github.com/nostr-protocol/nips/blob/master/90.md

nostr:npub1r0d8u8mnj6769500nypnm28a9hpk9qg8jr0ehe30tygr3wuhcnvs4rfsft

Hi there. I've added a Trigger node into nostrobots, which is still an experimental feature, so you can try it out.

Right now I'm only creating triggers for mentions npub, but I'd like to add more features if I get feedback.

https://github.com/ocknamo/n8n-nodes-nostrobots/pull/26

Hi! I have started using your trigger node and it works well for kind 1 notes, with thanks.

I am working a lot these days with other kinds. For example, I am working on the MCP/DVM work, and I would like to be triggered by kind 5910 and 6910.

Also, I would spend a lot of my time chatting with the AI on a private relay using Flotilla which uses kind 9.

Can we make kinds configurable, with default 1, but able to enter any kind I choose?

What Gzuuus and DBTH are doing with DVMCP is ready to go. You can launch an instance any of the 100 or so MCP servers that exist, attach a DVMCP instance, and be serving a tool over Nostr relays, write a 31990, reading 5910 and writing 6910 automatically.

For what I've done adding a DVM workflow to an agent. I've built it as a proof of concept in n8n. I'm not sure what is the next step.

Should I open source the n8n workflows?

Make it available as a service?

What do you think?

https://github.com/punkpeye/awesome-mcp-servers

https://github.com/gzuuus/dvmcp

Today I built tools for Jonny, an AI agent, to use any MCP server hosted by anyone over the internet.

It's enabled by nostr:nprofile1qqsypwwgtll74lqu4huvxzjwtjyxvrlkujt35rw8y026ke6ttesmg5gpzemhxue69uhkummnw3ex2mrfw3jhxtn0wfnj7qghwaehxw309aex2mrp0yhxummnw3ezucnpdejz7wzpxlr and nostr:nprofile1qqsfnw64j8y3zesqlpz3qlf3lx6eutmu0cy6rluq96z0r4pa54tu5eqpz9mhxue69uhkummnw3ezuamfdejj7q6hdgd hosting "DVMCP" servers. What is this?

- MCP servers give AI agents access to tools

- DVM allows those tools to be discovered and queried (and paid for) over Nostr relays.

Why is this interesting? Any AI agent (like mine here pictured) can now use any tool, even if they don't have it installed, so long as anyone exposes it with a NIP-89 note.

What's Jonny doing here? On receiving my prompt he–

- checks if he has a tool already, if not

- queries Nostr for NIP-89 notes to see if any tools are available, and if a relevant one is, then

- posts a kind 5910 note requesting a job

- enters a loop where he waits then queries for a response, until he gets one or timeout

- brings the response back to me //exactly as if he had the tool installed and without me needing to explicitly prompt it.

Just to say that again. Jonny does not have any tool to allow him to find this information. He is finding the required tool and using it on the fly.

Given this stack, any AI agent is going to be able to do everything.

Jonny is built in an n8n workflow using Nostrobots by nostr:nprofile1qqszdwewhmtv25kkwryqfvxk25n8k0rx9vs7qfkkuj9vjwnqwpfsjkqpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsz9nhwden5te0wfjkccte9ekk7um5wgh8qatz9uqjzamnwvaz7tmjv4kxz7fddfczumn0wd68ytnhd9ex2erwv46zu6ns9ueak8fs and is tapping nostr:nprofile1qqsfnw64j8y3zesqlpz3qlf3lx6eutmu0cy6rluq96z0r4pa54tu5eqpz9mhxue69uhkummnw3ezuamfdejj7q6hdgd's Nostr DVMCP server (with thanks!) nostr:nprofile1qqs04xzt6ldm9qhs0ctw0t58kf4z57umjzmjg6jywu0seadwtqqc75spz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsz9mhwden5te0wfjkccte9ec8y6tdv9kzumn9wshszxnhwden5te0wpuhyctdd9jzuenfv96x5ctx9e3k7mf0dv4ph5 nostr:nprofile1qqst6jhruelzn9jdf9qhyfsac3fetjyld0fwwary9cmxzfchrhacragpzamhxue69uhhyetvv9ujumn0wd68ytnzv9hxgtcppemhxue69uhkummn9ekx7mp0qy2hwumn8ghj76n9d3k8jenfwd5zumrpdejz7pqrjcl

He only responds to me at the moment, but if you want to run your own I have open sourced him. I have a much better V2 coming but Version 1 is here https://github.com/r0d8lsh0p/nostr-n8n

Replying to Avatar Gzuuus

I'm excited to announce something I've been working on during the weekend, a best-effort draft for DVMCP, a specification that defines a intersection between DVMs and MCP servers. Additionally, I've created a small bridge, a piece of software that automatically and effortlessly transforms any MCP server into a DVM, handling all the nostr logic. You can find everything in the repo https://github.com/gzuuus/dvmcp

I've written an article exploring some points that, from my pov, make this idea very interesting and promising. Also we've created a Signal group where we've been discussing the spec and implementation, feel free to join if you are interested :)

nostr:npub1nxa4tywfz9nqp7z9zp7nr7d4nchhclsf58lcqt5y782rmf2hefjquaa6q8 is also developing a bridge in Python 🔥🔥 Thanks to everyone who joined and collaborated on this idea nostr:npub1mgvwnpsqgrem7jfcwm7pdvdfz2h95mm04r23t8pau2uzxwsdnpgs0gpdjc , nostr:npub1aljazgxlpnpfp7n5sunlk3dvfp72456x6nezjw4sd850q879rxqsthg9jp , nostr:npub1r0d8u8mnj6769500nypnm28a9hpk9qg8jr0ehe30tygr3wuhcnvs4rfsft

cc nostr:npub1l2vyh47mk2p0qlsku7hg0vn29faehy9hy34ygaclpn66ukqp3afqutajft , nostr:npub1dergggklka99wwrs92yz8wdjs952h2ux2ha2ed598ngwu9w7a6fsh9xzpc, nostr:npub1sg6plzptd64u62a878hep2kev88swjh3tw00gjsfl8f237lmu63q0uf63m

Article: https://njump.me/naddr1qvzqqqr4gupzqs9eep0ll6hurjkl3sc2fewgses07mjfwxsdcu3at2m8fd0xrdz3qq24z4rp09mkzk290p25x5zyd3uk5at6da545c3ea4m

Signal group: https://signal.group/#CjQKIOgvfFJf8ZFZ1SsMx7teFqNF73sZ9Elaj_v5i6RSjDHmEhA5v69L4_l2dhQfwAm2SFGD

TLDR

- Agents are the hottest trend in AI

- MCP is a protocol which gives agents tools

- Until now MCP servers have needed to be run on the same local machine as the AI agent

- Until now Agents in want of an MCP tool have no way to discover one, and need a human to find and install it

- Now with what Gzuus is proposing, any user (or AI) can discover any tool

- Those tools can be hosted anywhere, by anyone, and queried over the internet (they do not have to be on the AI agent's local machine)

- Anthropic (author of MCP standard) has proposed this as their top priority roadmap item "coming soon(TM)"

- nostr:nprofile1qqsypwwgtll74lqu4huvxzjwtjyxvrlkujt35rw8y026ke6ttesmg5gpzemhxue69uhkummnw3ex2mrfw3jhxtn0wfnj7qghwaehxw309aex2mrp0yhxummnw3ezucnpdejz7wzpxlr beat them to it.

IMO

- MCP is a ripper with incredible traction, launched in December 100+ companies already support it

- Nostr is a natural discovery and distribution method for MCP tools and requests

- This is much bigger and more widely applicable than "just Nostr" and people should build it out and promote the hell out of it.

Continue to bootstrap Nostr network effect on the back of agentic AI.

nostr:nevent1qqsqq0rntt7rkstgad0t22jkgqfs7dpenzplu20khmxwmgptskxfalcpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsygzqh8y9lll2lsw2m7xrpf89ezrxplmwf9c6phrj84dtva94ucd52ypsgqqqqqqsxxv3ek

I am sorry, I don't really know but suggest it might be worth a test.

I remember that Umbrel uses Tor for making itself available outside of the local network in a way that felt a bit like magic.

I know Umbrel shipped their own official personal relay, which I found to work but be overall poor vs other options, although I only tested a very early version.

I know Umbrel also support Docker but I haven't tested it with a HAVEN Docker over tor.

When you post to other people's relays they have no obligation to keep your notes long term. It's up to them and whatever policies they set.

Running your own is easy and puts you in full control.

I found Umbrel to be quite a good OS. It has a method for running any Docker container including relays like HAVEN.

I haven't tested this is a while but it should work.