RYO ITABASHI

RYO ITABASHI · Creative Director

I don't write the code. I direct the AI — and nothing is called done until it works on real hardware.

Creative Director building Shelly, an open-source Android terminal and AI agent runtime, with AI coding agents.

See the Shelly case study Talks & project inquiries
Versions, over 8 months
v4.2.0 → v8.2.0
Commits on main
2,830
Versioned releases
21
Test device: Galaxy Z Fold6
N=1

As of 2026-11-04. Counts from git and GitHub Releases.

See it run

Shelly on an unfolded Galaxy Z Fold6, with terminal, AI, browser and Markdown panes in a 2x2 grid. Watch the demo videos on GitHub

The clips show Shelly running on the phone: Codex and Claude Code running on one phone in the demo, a write-and-review loop, and ChatGPT asking the phone to act, with the approval on the phone. Opens the Shelly README.

How I work

  1. 01

    Decide

    Me

    What the product is for, what goes on screen and what doesn't, which features to cut, which risks are acceptable, and when something counts as done.

  2. 02

    Implement

    AI coding agents (Claude Code, Codex)

    Native modules, the app, the runtime, tests, investigation and first-pass reviews. At high-risk points the agents review each other, and I decide which findings matter.

  3. 03

    Verify

    Galaxy Z Fold6 (N=1)

    A feature isn't done until it has been seen working on the phone. Whether a result counts as a pass is my call.

Read how this played out on Shelly →

Writing