pavo-etc

If me planning the requirements takes >1h the the POC is certainly the cheaper option

.... i'm hearing the solution in the question

Either way... still a lot of farmers

Can these companies please standardize their naming conventions for their tiers, I really don't want to learn another one.

haiku/luna = S

sonnet/terra = M

opus/sol/spark = L

mythos/fable/astra/wonder = XL

whatever the next bigger model is = 2XL

This is exactly me and I've fallen in love with nix

imagine wanting a device you can trust to not kill itself...

Glad that someone else can get some use out of it!

> explore all the CLIs Meta created for Muse to interact with stuff, like instagram or facebook (you can scp them off the VM)

oh this is cool. I would be very interested in a Facebook CLI

Fraud is generally considered shameful, not being an entrepreneur

I've put it into an XMPP wrapper for all communications so I can talk to my agents from any device supporting XMPP (all devices). This also means agents can talk to each other using group chats. I'm building apps from my phone on the bus, and handing off server admin tasks to the agents since they run in their own NixOS account with limited permissions.

Very similar to RMarkdown, though Ledge seems more flexible. https://rmarkdown.rstudio.com/

I really like this, this would definitely work for my brain. Might have to pitch it to some friends...

I have a harness built on top of pi that uses XMPP for all agent communication with me and between agents[0]. It works well, I text them from my phone and from my browser (or anything else that supports xmpp). They can communicate with each other over group chats and can schedule agent tasks using a GitHub project and a cron job.

The agents run on my NixOS server in their own unix account, this is the real secret sauce. This lets me use unix perms to limit their access to things, and it allows them to handle server administration. I've given them access to a cli mail and calendar client[1], github, and they have the ability to deploy updates whitelisted services. NixOS lets them install whatever they want in ephemeral shells, and it's not possible for them to break the server in a way I can't trivially roll back.

They can modify the caddy config, so there have been times I've entirely built and deployed new services on a new domain from my phone on the bus.

The harness is called pi-msg, I've done a writeup of my machine's setup on my site.[2]

[0]: https://github.com/zachpmanson/pi-msg [1]: https://github.com/zachpmanson/docket [2]: https://notes.zachmanson.com/the-fleet/

Just yesterday I ran a comparison of $/message through my harness[0] comparing Deepseek models to Luna, and to my surprise Luna won. I suspect its partially due to Deepseek's long thinking times, and also maybe due to OpenRouter variance in cache pricing etc.

Model and observed window | Messages | Retrieved actual $/message

DeepSeek v4-flash — all observed snapshots, 20 Jul–15 Sep | 4,610 | $0.0241

DeepSeek v4.1-flash — 11–27 Sep, before the 28 Sep billing change | 1,079 | $0.0576

DeepSeek v4.1-flash — 28 Sep, partial new billing window | 69 | $0.0365

GPT-6 Luna — 23–28 Sep, partial final day | 88 | $0.0329

I've subbed to Codex because I suspect at my usage rates the Codex Plus plan gives me more Luna messages than I'm using, and I've not really observed and better or worse intelligence performance. Interested to see how my $/message comes out after a month of usage on the Codex plan.

Something nice I've realised about my harness is that I can run different agents on different models so I can collect pricing data for a bunch in parallel.

[0]: pi-msg, run pi agents over xmpp https://github.com/zachpmanson/pi-msg

I have been using Mastodon since 2019 across multiple instances. The things you are talking about happen very rarely in actual use, I've never once had a problem with any of them. The signal to noise ratio on mastodon is higher than any other competing platform I've used.

This is extremely cool. Nix just continues to impress me more and more.

Whatever openrouter puts me on, running in pi (though I do all comms over my xmpp wrapper).

I'm not convinced spymark is better than just "invisible watermarks", spymark to my ears sounds designed to be sound very negative when invisible watermarks are not always negative, e.g. the counterfeit bank note example.

Tech like SynthID I see a net positive especially since it doesn't degrade text quality. I dream about a browser extension running at all times that makes text more translucent based on the confidence of LLM writing[0].

This article's suggestion of using it to unmask whistleblowers is very interesting and not something I'd thought about though. Still not convinced that spymark is a better name though.

[0]: Sean Goedecke's Deckard is close but it would rather invisible than bright red https://www.seangoedecke.com/deckard/

People can be multiple things.

I've done something similar by running pi within an xmpp wrapper on my server, so I can talk to it from any phone or computer. Works super well.

This in insanely cool.

Farid Zakaria's recent nix saga is insane

- unifying all versions of all packages into one resolution tree

- making executables out of sqlite databases

- adding API routes to a sqlite executable with an insert statement ?!?!?

- making one flake that contains all flakes

- adding version ranges to nix

- run any nix package any nix version in your browser (this one)

I would go even further, I want a browser extension that scans all words on every page and colours them more and more transparent as the likelihood of llm prose is increased.

Pi is a great example of this. I've used it as the basis for a fleet of agents I'm running on my server, but rather than using the Pi TUI I run it in headless-mode and have extended it to use XMPP as its communication layer so I can use any device with the same sessions, and so agents can talk to each other.

The Pi harness is built to be extended like this, and it's a joy to work with.

I've come to a similar lofi solution for my agent fleet. Markdown wiki with simple querying is decently effective as a memory system. Setting up a skill that can effectively reduce a session into useful long term lessons is the easiest unlock for these systems.

Luckily these are merely redirects!

Thanks! I had the same urge, I didn't want to be responsible for an entire GUI and communication layer but I did want to have a remotely controllable extensible harness, this was the perfect middle ground.

The benefit beyond that was having agents running on my server directly so they can do a lot of sysadmin and monitoring tasks, as well as suggesting config changes through PRs to my nix config.

Pii is interesting!

pi-msg doesn't support bonjour, I'm running it on a nixos server (nix is the real secret ingredient of my whole setup) the xmpp server is the same machine as the agents. I'm almost always on a different network to the one where the server is. If the server goes down the agents and ejabberd are coming down simultaneously.

I ran into this exact problem. I needed to fork Conversations to add a lot of polish. My fork is definitely improved but still has a few rough edges I'm chipping away at.