How To Design In The Agent Era

Video summary

Paper founder Stephen Haney says HTML/CSS-based design and human judgment can help teams ship faster without AI-generated sameness.

Paper founder Stephen Haney explains how the 12-person team built an agent-native design tool on HTML and CSS, letting agents edit the same code designers see and reducing format conversions, token use, and handoff friction. In demos, Paper creates shaders and image variations, imports live websites as editable layers, and sends designs back to code through tools such as Conductor and Cursor. Reviewing Legion Health, Sytex, and Moreta, Haney and YC’s Aaron Epstein flag common AI design tells: overly bold typography, too many font sizes, purple gradients, decorative cards, and clutter. Haney recommends limiting type to three sizes, reducing weight, and deleting elements that add no meaning. He argues agents can accelerate execution, but humans must supply taste and organizational decisions. Paper’s team reads every line of code; Haney says a brand designer built and shipped its website in about a week with agents. The company plans components, comments, and prototype feedback tools.

Chapters

1. 0:00Great companies have great design: Haney says human-led design helps companies stand out
2. 1:04What is Paper and why build it? Haney’s 12-person team builds an HTML-and-CSS design tool
3. 3:17What makes Paper agent-native: HTML and CSS link visual design directly to agents
4. 5:11Demo: Paper shaders, four-model image generation, and the upcoming Paper Mono font
5. 11:32Design-to-code and the new agent stack: Paper connects Conductor, Cursor, and the codebase
6. 16:48Design Review: Legion Health’s trust cues, live-site edits, and GPT 5.5 variations
7. 24:11How to avoid AI design slop: use three font sizes and lighter weights
8. 28:26Design Review: Sytex’s construction use case needs clearer above-the-fold messaging
9. 32:42The biggest tells of AI-generated design: cluttered cards, widgets, and purple gradients
10. 37:39Design Review: Moreta’s international QR wallet needs trust without losing energy
11. 41:11Can AI learn taste? Models may improve tactically, but humans still make the key decisions
12. 43:56How Paper uses agents internally: Agu shipped its marketing site in a week
13. 46:52Building a community around Paper: 25,000 early Twitter followers and traction alongside Sketch
14. 49:15Lessons from Stephen’s first startup: Radix UI, user feedback, and bottleneck-first decisions
15. 53:30What’s next for design and Paper: human-centered design and canvas-based prototype feedback

What the full analysis includes

15 chapters · 82 key points · 15 section summaries · mind map

The remaining 79 key points, section summaries and mind map are in the full analysis. Analyze this video to see everything →

This video is 57 min. The Free plan analyzes videos up to 30 min; plans from Plus cover videos up to 2 hours.