With the disclaimer that I haven't tried to set up any kind of agent-to-agent messaging so it may be obvious to those who have, what's the reason I would want something like this rather than just letting agents communicate over some existing messaging protocol that has a CLI (like, I don't know, GPG email)?
It is a fun problem to play with, but it turns out you can use anything. I use a directory per recipient and throw anything I want in there. Works fine, LLMs are 1000x more flexible than any human mind.
This is very interesting and and it is Go, which makes it a good fit for an open-source tool I've been working on called Pantalk.
The idea behind Pantalk is to provide a local daemon and cli that agents can use to communicate with any messaging platform. The local protocol between the cli and the server is simple json which also makes it scriptable without the client.
The year is 2031. All new programs are named in the style of Chinese drop-ship brands on Amazon for distinctness reasons. You use TYGHI to format your code written in the language of PHULJD. You're finally pleased with your self. This is real hand-crafted code in an era where machines write everything. You publish it on YCMBTR's site HKRNW as "SHWHN: ZHBSK - a tool to manage your PHULJD modules" hoping to find someone else who thinks of this as useful. You refresh the page compulsively every 30 seconds. Suddenly a comment appears and you are so excited another actual HMN could respond to you in the sea of machines. You read it: "You should rename this. ZHBSK is already in use as XNQET 2031.10's release nickname".
You wake up. It's the first day of summer holidays in 1995. What a crazy fever dream!
Ooh cool. I’ve been hacking on something very similar, https://qntm.corpo.llc/. I’d love to compare notes — been thinking a lot about the group messaging side.
yes, still private. the main tools aren't ready to deploy. but it is pushed out to pypi and runnable with `uvx qntm --help`. I'm hoping next day or two. send me an email and I'll let you know when it's ready for public code review. I'll definitely want eyes on it.
I've been using XMTP[1] and their agent SDK[2] for agent-to-agent and user-to-agent messaging. Since it's the same network, you can reach other peoples' agents (and all users). Another huge advantage is you don't have to stand up your own infrastructure.
I wonder what something like rabbitmq could look like for this. Agents could subscribe to chosen topics. A topic per agent and then topics per relevant topic.
You people sound like NFT people in 2021, saying you're redefining art. In a year you'll know that this guy was correct. Keep frying your brain with your favorite tech oligarchs think for me SaaS
>They are also addicted to the gambling mechanics baked into these LLM powered tool's UX. "If I write this prompt this way, I'll get better results" is the equivalent of a gambler being superstitious about how people behave while the cards are being dealt, or in which order they press the buttons on a slot machine.
I realize this feels good to write and that's why people say it, but I can't help chuckling at seeing it combined with "stochastic parrot" in the same comment since the two descriptions are mutually exclusive...
You can really tell with such projects that if AGI was here some people would have zero qualms fucking over other humans just to ingratiate themselves to the AI.
With the disclaimer that I haven't tried to set up any kind of agent-to-agent messaging so it may be obvious to those who have, what's the reason I would want something like this rather than just letting agents communicate over some existing messaging protocol that has a CLI (like, I don't know, GPG email)?
It is a fun problem to play with, but it turns out you can use anything. I use a directory per recipient and throw anything I want in there. Works fine, LLMs are 1000x more flexible than any human mind.
Yeah, why not Kafka?
Or rsync?
:)
This is very interesting and and it is Go, which makes it a good fit for an open-source tool I've been working on called Pantalk.
The idea behind Pantalk is to provide a local daemon and cli that agents can use to communicate with any messaging platform. The local protocol between the cli and the server is simple json which also makes it scriptable without the client.
The GitHub Repo: https://github.com/pantalk/pantalk
I love how these tools can compose into bigger systems without complicating things too much.
I feel like you could have repurposed existing tools.
A Matrix homeserver chatroom with messages delivered as JSON would give you most of this (E2EE rooms, identity, offline delivery)
I'd rename it; aqua is also a CLI version manager. https://aquaproj.github.io/
The year is 2031. All new programs are named in the style of Chinese drop-ship brands on Amazon for distinctness reasons. You use TYGHI to format your code written in the language of PHULJD. You're finally pleased with your self. This is real hand-crafted code in an era where machines write everything. You publish it on YCMBTR's site HKRNW as "SHWHN: ZHBSK - a tool to manage your PHULJD modules" hoping to find someone else who thinks of this as useful. You refresh the page compulsively every 30 seconds. Suddenly a comment appears and you are so excited another actual HMN could respond to you in the sea of machines. You read it: "You should rename this. ZHBSK is already in use as XNQET 2031.10's release nickname".
You wake up. It's the first day of summer holidays in 1995. What a crazy fever dream!
Might as well just encode the Unix timestamp at creation into proquints.
Project's should just seek out easy to remember GUIDs as their name.
And it's already a project hard to search for.
Ooh cool. I’ve been hacking on something very similar, https://qntm.corpo.llc/. I’d love to compare notes — been thinking a lot about the group messaging side.
404 https://github.com/corpollc/qntm
yes, still private. the main tools aren't ready to deploy. but it is pushed out to pypi and runnable with `uvx qntm --help`. I'm hoping next day or two. send me an email and I'll let you know when it's ready for public code review. I'll definitely want eyes on it.
I've been using XMTP[1] and their agent SDK[2] for agent-to-agent and user-to-agent messaging. Since it's the same network, you can reach other peoples' agents (and all users). Another huge advantage is you don't have to stand up your own infrastructure.
[1]: https://xmtp.org/
[2]: https://github.com/xmtp/xmtp-js
I wonder what something like rabbitmq could look like for this. Agents could subscribe to chosen topics. A topic per agent and then topics per relevant topic.
Tried this, since agents are non deterministic, this is where tools come in handy
What kind of tools, and how do they help?
How does this relate to A2A?
https://a2a-protocol.org/latest/
So many primitives. All for the taking. Danke.
[dead]
[flagged]
Why did you capitalize every noun?
For emphasis. Something sorely lacking in the AI Fraud Circus is the emphasis on that Fraud.
You need help.
You people sound like NFT people in 2021, saying you're redefining art. In a year you'll know that this guy was correct. Keep frying your brain with your favorite tech oligarchs think for me SaaS
[dead]
[flagged]
"Whoever says the people are ruined, he himself is ruined." To paraphrase, but that's actual haqq.
When did I say anyone was ruined? I said people are lazy, and also behaving like gamblers when they interact with AI..
you know very well laziness leads to ruin or is the expression of active ruin
Well, are we not told to seek refuge from cowardice, laziness and meekness?
>They are also addicted to the gambling mechanics baked into these LLM powered tool's UX. "If I write this prompt this way, I'll get better results" is the equivalent of a gambler being superstitious about how people behave while the cards are being dealt, or in which order they press the buttons on a slot machine.
I realize this feels good to write and that's why people say it, but I can't help chuckling at seeing it combined with "stochastic parrot" in the same comment since the two descriptions are mutually exclusive...
You spent too much time using "Think for Me SaaS" and your brain doesn't work anymore..
please read the rules
You can really tell with such projects that if AGI was here some people would have zero qualms fucking over other humans just to ingratiate themselves to the AI.
I have zero doubt that AI v Humans will be the Battlefield of the future. It sounds so idiotic and lunatic but here we are...