I'm exploring using XMPP as a coordination layer for multiple LLM agents.

I'm using Herdr to run multiple agents in parallel, and my goal is to see whether I can get them to collaborate more efficiently.

For example, when one agent's PR gets merged, other agents immediately become aware of it, rebase their own PRs and fix conflicts.

What also often happens is that new tangential requirements and sidequests often arise while working in a session, and I now let agents hand these off to parallel Herdr tabs. The handoff can be done with or without XMPP. These parallel agents announce their statuses in the XMPP MUC (groupchat), which enables them to coordinate.

Technical details:

Converse XMPP client runs as a headless NodeJS daemon with a CLI that can be called by agents.

I created a /coordination skill which is enabled by default via a Claude Code hook. When an agent starts, it ensures the Converse daemon is running and then joins a password-restricted and members-only MUC where all the other agents in that Herdr workspace are as well (and also myself). I can then assign that agent tasks via messages in XMPP (DMs or groupchat) or from Claude Code.

Once the agent has the task, it announces it in the MUC so that other agents are made aware. When sidequests or requirements arise in that session, it can ask in the MUC whether other agents are idle and able to take over that parallel task.

I've been building the Converse daemon and CLI and have now been experimenting with the workflow and working out various kinks and use-cases.

There are various other XMPP features that can be explored further. I'm thinking of using PubSub with a web UI with vertical swimlanes which can show me which agents are currently active, what they're working on and for how long.

I'm using Claude Code because that's what we use at $WORK, but this should be independent of any harness and part of the point is create neutral tools not locked into any particular inference provider.

This would be more convincing if it wasn't from a site trying to sell me gold. Do people really believe that the mechanization of clothing production in the industrial era has made no difference to "real" prices?

(Rome definitely had inflation crises!)

Those are credit bubbles bursting, not the result of hard money.

This is a almost entirely oversimplified take on the Great Depression to the point of meaninglessness

For a decade before Black Thursday,there had been many things that were signs that the economy was having trouble even if the "Roaring Twenties" made it seem like everything was fine.

IMO the largest issue was that American farm sector was teetering on the edge because of the dramatic drop in crop prices. This deflation screwed over farmers who mechanized with lots of debt, which because of said deflation, became impossible to pay off.

The fed also implemented rate hikes to curb speculation right before 1929 which froze up credit contributing to deflation

The problem of the Great Depression was NOT the stock market crash, it arguably wasn't even the real start, just the most "spectacular" one. The problem was that with the entire economy deflating, it caused a massive downward spiral that the Fed did not really have the tools to fix, because of Gold Standard and lack of legal authorization.

This was why the Govt went to extreme lengths to try and figure out how to raise prices, which is why you get programs to pay farmers to NOT grow food, and mass killings of pigs and cows and other farm animals, even as the farmers who raised those lifestock went hungry.

So no, speculation was not the problem, it just sparked the key issue of the fact that the economy was deflation uncontrolled, but was just hidden.

This is a complex topic and I think you have done a good job of summarizing the main issues. To add a little context:

>...that the Fed did not really have the tools to fix, because of Gold Standard and lack of legal authorization.

This was just bad policy by the Fed. The Fed had the legal authority to be the lender of last resort and could have prevented the bank failures. Many explanations have been given over the years as to why the Fed didn't provide liquidity. Because the Fed failed to supply emergency liquidity, the U.S. money supply plummeted by nearly 30% over the next couple of years, which essentially turned what likely would have been a recession into the Great Depression.

This is not to say the gold standard was not a problem. During the 1930's, leaving the gold standard was one of the few good moves done to help the economy by the federal government.

Yes, this often happens in pairs: the overcorrection after excessive credit results in a deflationary money market, destroying any chance at recovery.

FWIW Converse.js and Dino also support the latest OMEMO.

Oh, nice! It looks like Converse.js has added it just two months ago, but I haven’t been able to find anything for Dino (only complaints about as old as the latest release, about the newest OMEMO not being supported). Was it shipped very silently, or are they perhaps planning to add it in an upcoming release?

Looks like I might have been wrong about Dino, sorry about that.

I worked on adding OMEMO2 support to Converse.js via libomemo.js and I tested it against Kaidan.

Generally speaking, if demand stays high, and existing suppliers don't increase supply, the high prices create an incentive and opportunity for new market entrants to increase the supply.

Yes that would be my exact point.

Yes, "quickly" is obviously industry-dependent. If they were fabricating something extremely easy to manufacture, "quickly" might mean months-to-a-year. Fabricating something very complicated like memory "quickly" might mean years.

That said, even though we might have to wait years for new supply to hit there market, we probably don't have to wait years to see signals of that. We will likely know that manufacturers are starting to build new supply long before we see the chips hit the market.

In the current environment unless people start decommissioning datacenters en masse supply will be locked up till the early 2030s IMO.

Even if you started a new fab today, it's 5 years till it comes online. The great hope here is that China's DDR5 yields represent a high reject rate of an enormous amount of new capacity that they'll dump into the market to get a sovereign capability in that area.