To build on this, I think having a strong VC system for open models to flourish and be optimized for different use cases, while having communities contribute to their development could be key for open-weight RSI.
@darrenjr Just you wait until @CoHordeAI has its beta. We're gonna have a pretty cool interface for agents to say the least. In terminal and outside of terminal
@michael_chomsky I think Hermes has a lot of potential, but I worry that it could get burned by not having a good orchestration of tasks. My setups have always seemed to be a bit pricier than I think is possible
@imbabybrooklyn Super interesting! I love this UI. We're experimenting with wide variety of fine grained and abstract workflow creation at @CoHordeAI
Would love to connect!
Definitely a super exciting launch! We agree that agents wrapping around compute and working together is the future of development. It's been speeding our work up already!
We're taking this a step further and creating custom orchestrations of multi-agent systems at a per-task level.
Excited to see how QM evolves and how we can work alongside it!
We’ve decided to open-source a multi-agent harness we use internally at YC.
We call it “QM” and it’s meant to be easy to customize, like Hermes or OpenClaw, but useful for a whole company. We use it across accounting, legal, events, and engineering (including building QM itself!).
The whole project is under an MIT license. It is cloud-first and has Slack and web UI natively.
Congrats on shipping! QM proving harnesses are interchangeable matches what we noticed too, we also wrap Claude Code and Codex behind one dispatcher. Which means the interesting layer is now above the harness: decomposing a goal into a DAG of small, purpose-bound agents so you control cost and blast radius per step (aka powerful orchestration). One agent in a workspace is a chat, but a graph of them, each with a job description and a budget, is a workforce. Would love to swap notes! Been saying for a while that the future is multi-agent driven 😉
applying to @ycombinator is honestly just a good forcing function. whether we get in or not, the form itself changed how i see my startup. here's some things i noticed:
> you realize you can't explain the product without buzzwords
> listing what you shipped this month exposes your real speed
> trying to articulate why incumbents won't crush you exposes how fragile your moat might be
> you're forced to admit how much of the founding team can actually build the thing
> answering what happens if it fails proves if you really care about the problem
dropping the delusion for a bit makes it super worth it. gl to everyone submitting today!
One of the problems @CoHordeAI is trying to solve. I know CS majors that still copy paste code from Chat. Agents are the future, and will allow us to focus on creativity and thought. So we're trying to bring that to everyone!
@liam_fallen Typing lets me be insanely targeted and intentional. I find myself doing a stupid amount of editing after voice prompts
Oh and I mobile prompt at the gym and work so there's that
@sflorimm I think people would start optimizing for cost. Thinking about their providers and orchestration layers a lot.
Prices are gonna go way up from frontier labs, but token costs overall will go down with quantized and open-weight models. Routing is gonna be super important
I do not miss the days of
"You question doesn't really make sense if you're using C correctly."
"Are you sure you need to be doing this?"
"This is something you should've learned right away in this language."
"Maybe you should focus on learning the basics before you try taking this on."