Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Finally. I've been saying that we need to stop focusing on a single agent getting everything right and instead layer agents for about 16 months now, but it's great to have a paper to point to.

It's interesting that the diminishing returns for tasks flatten out rapidly around the same size as the ideal human meeting sizes: https://www.researchgate.net/figure/18-Optimal-Meeting-Sizes...

If this was done at more granular steps of agent quantity I'm curious just how closely it would match those numbers.

I'd also really love to see the eventual follow-up where we see how much more performance can be obtained when the agents are each fine tuned towards slightly different aims. I'd expect there'd even be a performance lift from just having the agents each set at different temperature levels.

Very happy to see the research community starting to step in this direction!



I couldn't agree more. You should check out LLMWare's SLIM agents (https://github.com/llmware-ai/llmware/tree/main/examples/SLI...). It's focusing on pretty much exactly this and chaining multiple local LLMs together.

A really good topic that ties in with this is the need for deterministic sampling (I may have the terminology a bit incorrect) depending on what the model is indended for. The LLMWare team did a good 2 part video on this here as well (https://www.youtube.com/watch?v=7oMTGhSKuNY)

I think dedicated miniture LLMs are the way forward.

Disclaimer - Not affiliated with them in any way, just think it's a really cool project.


Great videos.

I have one personal niggle: I get annoyed when we end up lying to ourselves. Regarding the 101 section in video 1 - People forgot this the day LLMs came out. I felt this was too generous with the benefit of doubt.

This basic point was and remains constantly argued - with “Emergence” and anthropomorphization being the heart of the opposing argument.


This is how I think humans work. We have 5 or 8 versions of us running around in our skulls or whatever and one of them is somewhat of a supervisor.


We have tons of specialized components that work together cooperatively and competitively. There’s multiple ways they connect. There also seems to be global processes that happen, like during sleep. There’s over 3,000 cell types per the BRAIN initiative. Every brain forms on it’s own taking shape like something out of a Transformers movie.

God’s design is mostly nothing like man’s neural networks. It’s far superior. Brains are also what’s creating all the artificial, neural nets on top of all math, tech, and economic systems that they run on. AI’s got a lot of catching up to do.


I think it's way more than 8 even. And it's common to have many working as supervisors, often at conflict with each other. And some act out the automatic trauma responses, as they're stuck in the past when the trauma occurred.


This sounds very much like the internal family systems model: https://en.m.wikipedia.org/wiki/Internal_Family_Systems_Mode...


Right it’s like a random forest or something maybe.


minsky, 'society of mind', everything old is new again


Multi Homunculus Model


I was thinking something like Julian Jayne’s Bicameral mind, but you’re always arguing with yourself lol


Sounds like dennett's multiple drafts hypothesis.


I personally visualize them at a table from time to time.


Do they have to collaborate to decide who gets to use the forks?


That sounds like a culturally bound phenomenon, shouldn't they be using chopsticks?


Chopsticks makes sense because there's only 5 of them and they have to share the 5 chopsticks.


Of course not, you don't need forks while playing poker.


How do they look? Do they look differently?


Maybe. I'm sure one's consciousness corresponds with one's guiding philosophy.

I don't think this supervisor model is generally applicable to people with EFD or some forms of Autism, for example.


And et voila, you have the script of inside out. \s

But honestly I do think this is how we operate. Depending on our state of metabolism and other psychological factors, the dominant version changes but as a whole we remain the sum total of all these versions.


So there is a number, let's say 42, 42 versions of us running in parallel make consciousness, reasoning and all other useful abilities appear?


More like a number within Miller's law[1].

https://en.wikipedia.org/wiki/Miller%27s_law#:~:text=The%20o...


I was working on multi-agent systems for problem solving via https://github.com/agi-merge/waggle-dance for months last year!


“Each fine tuned toward slightly different aims”

So…a sort of mixture of experts if you will


Kind of. More like a mixture of a mixture of experts.

The problem is MoE on its own isn't able to use the context as a scratch pad for differentiated CoT trees.

So you have a mixture of token suggestions, but a singular chain of thought.

A mixture of both is probably going to perform better than just a mixture of the former, especially given everything we know by now regarding in context learning or the degree of transmission synthetic data is carrying.


It seems funny that the researchers are studying what people are building to experiment. crewAI is one example.


I've had a POC agent-chain thing written in haskell for about 9 months. this paper is a little different though.




Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: