How To Design In The Agent Era
Stephen Haney, founder of Paper, discusses how his AI-native design tool uses HTML/CSS rendering to enable seamless collaboration between designers and agents. The conversation explores design fundamentals, common AI-generated design pitfalls, and how Paper is disrupting the design tool market by prioritizing human creativity while accelerating workflows with agents.
Summary
Stephen Haney presents Paper, an AI-native design tool built by a 12-person team of designers and engineers. Unlike traditional design tools with custom rendering engines, Paper uses HTML and CSS as its rendering engine, making it inherently compatible with AI agents that understand web standards. This architectural choice results in faster execution, lower token spend, and higher accuracy when agents interact with designs.
The conversation covers Paper's technical advantages and design philosophy. Because Paper renders HTML/CSS natively, designers unknowingly create code-ready designs that agents can understand directly. The tool includes innovative features like customizable shader effects (created by studying physical paper's randomness), image generation libraries using multiple models simultaneously, and direct integration with code repositories through MCP servers. Stephen demonstrates how designers can extract colors, vectorize graphics, and export designs as React components or Tailwind CSS.
A major focus is identifying and avoiding "AI tells" — visual patterns that signal a design was generated without human refinement. Common tells include: excessive font weights (designers should pull back to regular or light weights), too many font sizes (consolidate to three), overuse of cards and widgets, unnecessary icons and decorative elements, side swoops of color, purple gradients, dark mode implementations, overly spaced all-caps headers, and numbered badges. Stephen argues these patterns don't inherently look bad but become problematic through repetition, making designs look generic and untrustworthy.
The discussion emphasizes that human design judgment remains irreplaceable. While models like Fable show improved taste and thoughtfulness, they cannot make high-level organizational decisions about what to build, which stakeholders to satisfy, or how to communicate value propositions. Models excel at tactical improvements (typography, spacing, contrast) but struggle with strategic design thinking. Stephen argues that exceptional design is the differentiator between great companies and average ones, citing that every successful company of the past 20 years has had exceptional design.
Paper's go-to-market strategy relied heavily on values alignment before the product existed. Stephen spent the first year building Twitter followers and podcast presence by authentically discussing design principles. This grassroots approach created a movement among designers who appreciated the team's genuine passion for design tooling. The company has achieved comparable adoption to Sketch (according to Ramp subscription data) within a highly competitive market dominated by Figma.
Stephen contrasts Paper with his previous company, Modules, which attempted designer-developer handoff but failed to achieve venture-scale business. He emphasizes learning to focus on "bottlenecks first" — identifying what's preventing growth and solving that rather than building comprehensive feature sets. At Paper's current stage, the bottleneck has shifted from awareness to missing features like comments and components, which will guide development priorities.
The conversation demonstrates Paper in action across multiple use cases: creating animated shader effects for YC branding and acceptances, generating brand illustrations with multi-model exploration, extracting color palettes, live editing websites captured via Chrome extension, and requesting agent iterations on designs. Real website reviews (Legion Health, SciTechs, Moretta) show how pulling back font weights, reducing font sizes to three maximum, and removing decorative elements dramatically improves professionalism and trustworthiness.
Stephen's team internally takes a measured approach to agents despite building tools for them. All code receives human review from multiple team members because design tool development requires precision and consistency. However, agents assist with marketing sites, videos, and non-critical features. Agu, Paper's brand designer, built the entire marketing website in one week using Paper and agents, demonstrating designer-agent collaboration at its best.
About this episode
<p>AI isn't just changing the tools designers use. It's changing how they build, ship, and stand out. </p><p><br /></p><p>In this episode of Design Review, Stephen Haney, founder of AI-native design tool Paper, joins YC General Partner Aaron Epstein to demo the agent-first workflow that's making Paper one of the fastest-growing design tools since Figma. </p><p><br /></p><p>Using live redesigns of user-submitted websites as examples, they break down the most common AI design tells, show how to fix them in seconds, and explain why the biggest risk for founders isn't moving too slow, it's shipping something that looks like everyone else.</p>
Key Insights
- Stephen argues that using HTML/CSS as Paper's rendering engine creates lower token spend, faster agent execution, and higher accuracy compared to tools with custom engines, because agents already understand web standards from their training data.
- Stephen contends that exceptional design is the primary differentiator between great companies and average ones, noting that every successful company in the past 10-20 years has had exceptional design, making it essential for founders to invest care in design rather than accepting default outputs.
- Stephen identifies that models consistently produce designs with excessive font weights, too many font sizes, unnecessary cards and icons, and purple gradients — patterns that individually aren't bad but collectively signal a design wasn't refined by humans, undermining trustworthiness.
- Stephen argues that human designers will remain essential because design involves organizational decision-making, stakeholder management, requirement collection, and value proposition communication — tasks agents currently lack skill at.
- Stephen demonstrates that Paper achieved 25,000 Twitter followers in its first year before launching a product by authentically discussing design values, suggesting that values alignment with a market allows authentic products to generate grassroots support despite missing features.
- Stephen claims that consolidating fonts to three sizes and pulling back all font weights to regular or below immediately makes AI-generated designs appear more intentional and professionally designed with minimal effort.
- Stephen reveals that despite building tools for agents, Paper's internal code review is entirely manual with every line reviewed by multiple humans because design tools require precision that agents cannot yet reliably deliver.
- Stephen asserts that design's function in organizations — exploring problem space, competitive analysis, stakeholder communication — is not changing with AI, and the real need is helping designers keep up with engineering velocity to ensure quality ships with software.
Topics
Transcript
The agents can speed you up. They can save you a lot of work, and they can do tasks like translation or resizing for you. But we're big believers in the human element of design and how important that is. And so we're trying to create a tool that lets humans work really fast with agents. It's probably very tempting to use whatever cloud design spits out. I don't think it helps you stand out from the crowd. And so I think design is this great differentiator. And if you look at every great company of the last 10, 20 years, basically all of them have exceptional design. And so I think if you want to be one of those…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Y Combinator Startup Podcast
Garry Tan: Own Your Intelligence
Garry Tan argues that personal AGI—intelligence owned and controlled by individuals rather than rented from corporations—is already emerging through AI agents powered by personal context libraries and skill files. He draws parallels to Spinoza's fight for intellectual freedom to advocate for developers and founders to own their cognitive tools and infrastructure before platforms extract their expertise.
Building the First Data Centers in Space
Philip Johnston, CEO of StarCloud, discusses how the company is building data centers in space to solve Earth's energy bottleneck for AI by leveraging unlimited solar power and declining launch costs. The company launched its first satellite in November 2024 with high-power NVIDIA GPUs and plans to deploy thousands of satellites providing 20 gigawatts of compute capacity.
Waymo Co-CEO Dmitri Dolgov: "Move Fast And Ship Safely"
Waymo Co-CEO Dmitri Dolgov presents seven technical lessons learned from building and deploying autonomous vehicles at scale, emphasizing the critical difference between demos and production systems, the importance of architectural choices aligned with safety requirements, and the necessity of a comprehensive AI ecosystem including simulators and rigorous evaluation frameworks.
Patrick Collison: "What If You Succeed?"
Patrick Collison, CEO of Stripe, discusses his journey from dropping out of MIT to founding Stripe, emphasizing the importance of grounding decisions in real customer problems and the value of deep knowledge even in an AI era. He shares data showing unprecedented growth in new business formation and argues that this is the best time ever to start a company, with AI creating more decentralization rather than concentration.
Jeff Dean: The 1% Rule for Building in AI
Jeff Dean discusses AI progress, hardware specialization, and the future of agent-based systems. He argues that inference optimization and context engineering are key frontiers, while emphasizing that taste and problem selection—not just technical execution—will differentiate successful AI-native founders from competitors.