Superpencil
Visual-in, visual-out.
An agent-native visual thinking app. You draw, annotate, and sketch on an infinite canvas — AI agents respond visually in the same space. No prompts. No text boxes. The pencil is the interface.
We build the layer between AI systems and human intent: interfaces where humans use AI, and agents that learn how to reach humans.
Proof systems
Visual-in, visual-out.
An agent-native visual thinking app. You draw, annotate, and sketch on an infinite canvas — AI agents respond visually in the same space. No prompts. No text boxes. The pencil is the interface.
AI-to-human distribution intelligence.
A proof system for the last mile of autonomous software: AI reads public market signals, learns which humans can move belief, and turns that signal into structured creative direction. If Superpencil is humans using AI, Superdirector is AI learning how to use human distribution.
Acceleration
01
Superworld (ST World)
2 years
Manual Coding
20K users, featured on a16z market map
02
Superpencil
6 months
LLM Coding
First online-syncable raster drawing engine
03
Superdirector
6 weeks
Agentic Coding
Brand growth video intelligence product
Team
Founder & CEO
Bell prefers slower LLMs — he can't read as fast as they output. When he tries to describe a design idea in text, the idea dies in translation. So he started drawing things out on paper and photographing them as input. That's visual-in visual-out — he was doing it by hand before building it into a product. Coding since 14. Image generation paper at 18 (IEEE oral). Built Superworld (ST World) — 20,000 users, featured on a16z market map. Built Superpencil in 6 months, Superdirector in 6 weeks.