Sawyer Hood: Building AI's Weird Future in Public

Sawyer Hood represents a new breed of engineer: one who left prestigious roles at Figma and Facebook to pursue a single obsession—exploring every weird way humans can interact with LLMs. He builds in public, ships fast, and shares everything on Twitter.

The Philosophy: Experiments Over Products

Hood describes his current focus simply: “I’m a software engineer who is currently exploring all of the weird ways we can interact with LLMs.”

This isn’t about building the next billion-dollar startup. It’s about rapid experimentation—finding the boundaries of what’s possible when you give AI new interfaces to the world.

Key Projects

draw-a-ui (13.6k+ stars)

Hood’s viral hit: draw a wireframe mockup with tldraw, and GPT-4 Vision generates working HTML with Tailwind CSS.

The implementation is beautifully simple:

1. Take the canvas SVG
2. Convert to PNG
3. Send to GPT-4 Vision with instructions to return a single HTML file

"This works by just taking the current canvas SVG, converting it to a PNG, and sending that png to gpt-4-vision with instructions to return a single html file with tailwind."

No complex pipelines. No over-engineering. Just the shortest path from idea to working code.

Link: https://github.com/SawyerHood/draw-a-ui

dev-browser (3.3k+ stars)

A browser automation plugin for Claude Code that gives your AI agent the ability to control a web browser. Key features:

• Persistent pages: Navigate once, interact across multiple scripts
• Flexible execution: Full scripts when possible, step-by-step when exploring
• LLM-friendly DOM snapshots: Structured page inspection optimized for AI understanding

This solves a real problem: coding agents need to verify their work by actually using what they build. dev-browser gives them eyes and hands in the browser.

Link: https://github.com/SawyerHood/dev-browser

Terragon Labs

A remote background agent orchestrator for running Claude Code, Codex, and other coding CLIs in the cloud. When your AI agent needs compute, Terragon provides it. Now open-sourced.

Link: https://github.com/terragon-labs/terragon-oss

windows9x

An AI nostalgia trip: Windows 95/98 UI where all the apps are AI-generated. Pure experimental fun.

Link: https://github.com/SawyerHood/windows9x

Why This Matters

Hood’s work illustrates a key insight: the most powerful AI applications often come from giving models new ways to perceive and interact with the world—not from prompt engineering or fine-tuning.

Draw-a-ui gives AI vision of your sketches. Dev-browser gives AI control of browsers. These are new modalities, not incremental improvements.

Philosophy Takeaways

1. Build in public: Share experiments, not just polished products
2. Shortest path: SVG → PNG → GPT-4 Vision. No unnecessary complexity
3. New modalities > better prompts: Give AI new interfaces to the world
4. Weird is good: The best experiments are often the strangest ones

Resources

• Website: https://sawyerhood.com
• GitHub: https://github.com/SawyerHood (78 repos, 827 followers)
• Twitter: https://x.com/sawyerhood (primary channel for experiments)
• Email: sawyerjhood@gmail.com

Career Path

• Facebook (previously)
• Figma (previously)
• Independent builder (currently)

“I’m currently building all of my experiments in public on Twitter. If you are working on something cool or just want to chat, feel free to reach out to me there!”