Aug 7, 2026, 10:00 AM ETInterview, Product Demonstration, Conference Presentation

How To Design In The Agent Era

Paper is an AI-native design tool built by a team of 12 designers and engineers, led by founder Stephen Haney (formerly of Modules/Redix).
Core Philosophy: The tool is built on the belief that while agents accelerate speed, the "human element of design" remains the critical differentiator for standing out in a crowded market.
Adoption: Paper is currently used internally by Y Combinator (YC) for all design work and has attracted significant traction among founders, with data showing it is now the third most-used design tool behind Figma and Sketch based on Ramp credit card data.
Strategic Focus: The team prioritizes quality and precision over rapid feature accumulation, contrasting with competitors who prioritize speed and breadth.

Rendering Engine: Paper uses HTML and CSS as its native rendering engine to maximize agent compatibility, as LLMs have extensive training data on these languages compared to proprietary binary formats.
Agent Efficiency: Using HTML/CSS reduces token spend, increases execution speed, and minimizes hallucinations compared to tools requiring visual-to-code translation.
Visual Interface for Agents: The product functions as a visual interface for agents, allowing direct manipulation (drag and drop) rather than relying solely on text prompting.
Desktop & MCP Integration: A desktop application includes an MCP (Model Context Protocol) server, enabling agents to interact with local repositories and the design canvas directly.
Live Code Handoff: Designers can copy frames as React components or Tailwind CSS, allowing the code to be the single source of truth that syncs between design and development.
Chrome Extension: A browser extension allows users to grab live website sections and paste them directly into the Paper canvas as editable layers.

Paper Shaders: An open-source library of WebGL animations (e.g., loading states, halftone dots, neuro noise) that allows designers to create complex visual effects without coding.
Image Generation: A multi-model library (including a "variety pack") that allows simultaneous generation across four different AI models to explore creative directions.
Paper Mono: A custom typeface designed specifically as a versatile all-rounder for marketing and code, launching with a focus on legible numbers.
Vectorization & Color Extraction: AI tools that can vectorize images and extract color palettes from generated graphics for rapid iteration.
Prototyping Organization: Upcoming features will allow users to iframe external prototypes onto the canvas to centralize feedback and management.

Source of Truth: The tool treats the codebase as the source of truth to eliminate the maintenance burden of keeping design systems and code in sync.
"Vibe-Coded" Aesthetic Critique: The team identifies common AI design flaws to avoid, including:
- Overuse of bold font weights (recommended limit: light/regular).
- Excessive font sizes (recommended limit: 3 distinct sizes).
- Overuse of cards, gradients, and unnecessary icons.
- Poor contrast and alignment issues (e.g., text overlapping cards).
Iterative Curation: Workflows involve prompting agents to generate hundreds of variations overnight, after which humans curate, combine, and refine the best elements.
Comment-Driven Workflows: Users can leave comments on design frames for agents to address, allowing for asynchronous feedback loops that run overnight.

Founder-Market Fit: Haney started the project as a solo founder because of his personal passion for building creative tooling, prioritizing intrinsic motivation over market trends initially.
Previous Venture Lessons: Lessons from his first company (Modules/Redix) taught him to avoid trying to please both designers and developers equally, and to focus on solving a specific bottleneck rather than building a generic solution.
Community Building: The team grew to 25,000 Twitter followers and launched the product without traditional marketing by being values-aligned, authentic, and engaging directly with the design community on social media.
Internal AI Usage: While Paper's core engine is human-built for precision, internal teams use AI for marketing assets, code generation, and PR reviews, though all code is manually audited by humans.
Human-in-the-Loop: Agents are viewed as accelerators for tactical execution (resizing, translation, variations) but not as replacements for high-level decision-making, stakeholder management, or defining the "problem space."

Table Stakes Features: Immediate development focus is on adding components and comments, which are currently identified as the primary bottlenecks for user adoption.
Agent Ecosystem: The goal is to position Paper as the visual coordination layer in the "new agent stack," pairing with tools like Cursor, Conductor, and GitHub.
Trend Evolution: Haney anticipates that while AI will improve at tactical design rules, the fundamental human need to push visual trends and make high-stakes decisions will remain essential for at least the next decade.
Scalability: The team acknowledges the challenge of scaling server capacity to handle the volume of iterations generated by autonomous agents.