Yesterday our co-founder Jay Ma (@Doudoujay) asked a room of founders: has a post or a launch ever gotten you more attention than customers? Every hand went up. Then he asked them to keep it up if they could put a number on what that attention did to revenue. Most of the hands came down. We know the feeling. We were building @finalroundai before @hellyeah_ai, and it got to $1M ARR on viral alone. One post did 83 million views. Then we went after $10M and kept getting the views without the revenue. You can't schedule a viral post or buy another one. So we put paid next to it and scored every channel on what @stripe collected. $15M ARR in 12 months. 4.2x blended ROAS. @hellyeah_ai came out of that. We were customer zero, then other founders asked us to run the same playbook on their accounts. The first thing we find is broken measurement: 👉 One campaign we audited reported 2,757% ROAS on the platform. Stripe, net of refunds, said 48%. Same campaign. 👉 A client's dashboard counted 16 sales last month. Their database held 3. 👉 A streaming client paid $300 a sale bidding on purchases. 90% of their payers hit 12 streams on day one, so we moved the bid to that event and paid $80. We fix the data before anyone touches a bid. We got tired of doing this by hand on every account, so we built it into a product. The agent runs the audit and the budget moves, and it asks you before it spends. Google Ads today, more platforms after. Jay got into it with @unifygtm (@DurshtE ), @shanda_group (@bei_zhang01) and Press Club (@kwangaroo) afterward about GTM and which users are worth chasing in the first place. All three of us sell to founders who need customers to show up.

Today we’re launching a new, brief-first Creative Studio in @hellyeah_ai. Most AI creative tools make you start with a reference image or a blank prompt. Creative Studio starts with one brief and turns it into a complete ad set grounded in your brand. Here’s how it works: 1. Write your campaign brief 2. Choose from 5 messaging angles 3. Pick a visual style 4. Generate multiple creative directions 5. Choose the set you want to launch Each set includes the copy, a landscape image, and a square image. That’s it. One brief to a launch-ready creative set.

.@ComfyUI community helps build the product. In this conversation, @jojodecayz explains how the loop works. The core team improves model inference. Developers turn those improvements into custom nodes and integrations. Creators combine them into workflows that others can reuse and extend. Each channel supports a different part of the process: @X showcases finished work. @Reddit and @discord host technical experiments and troubleshooting. Templates and examples help new users get started, while in-person workshops help professional users learn the system with guidance. Together, these contributors give the next builder a larger set of tools to work with. A closed-source AI company would need to recreate that participation across the full stack, from model inference to production workflows.

ComfyUI’s new Local MCP makes more sense after hearing Jo Zhang (@jojodecayz) explain how the team thinks about agentic creation. The update lets an agent work with a user’s real @ComfyUI setup, including installed nodes, custom nodes, and models on disk. In this conversation, Jo explains the design choice behind it. Agents need access to the codebase and local assets before they can operate a creative workflow. Creators still decide what to make and how the output should look. Jo also explains why ComfyUI is building toward an operating system for visual creation. If you build AI agents or creative tools, this conversation is worth watching.