Robin Huang

Most of what earned @ComfyUI a $500M valuation today was built by people we don't employ. We just closed $30M, led by @craft_ventures with @PaceCap, @chemistry, and Tru Arrow. Total: $47M. Where it goes: → Comfy Cloud → Collaborative workflows → Comfy Local → Ecosystem stability → Day-one model support We're not building a walled garden. We're building open infrastructure, the kind that lasts and empowers rather than locks in. This started as a community project. 4M users and $47M later, it still is. If you haven't built something in ComfyUI yet, come find us: comfy.org

If you've spent the last two years building workflows on ComfyUI, you already know what these frameworks are asking you to do.

The word changed before the work did.

Every "agent framework" I look at in 2026 is a node graph with new branding.

Shoutout @github for being my most useful PM tool

Every pull request is someone who cared enough about a problem to write the fix themselves. That's a stronger signal than any user research I've ever read.

I open GitHub before I open analytics every morning.

Our team moves quick and you'll work directly with myself and the other founders on everything you ship. jobs.ashbyhq.com/comfy-org/9e4b…

You'll read source code when something breaks, ship before the spec is finished, and sit with users running production pipelines before deciding what to build next.

@ComfyUI just raised $30M at a $500M valuation and we're hiring our next product manager

Also native in ComfyUI this week: Google DeepMind's Gemma 4 and BiRefNet. blog.comfy.org/p/c237b7ce-29d…

This pattern is becoming routine. Major labs open-sourcing what would have been their moats two years ago.

Netflix just open sourced a video AI model that should have been their competitive moat.

This is what ComfyUI actually is. Not the codebase, but the people who show up when someone needs help.

They came back a few days later. Had written two custom nodes and fixed the whole workflow.

We have 188,000 people in our @Comfui subreddit. Someone posted a question last week: "Is there a trick for repetitive task?"