I've spent 12 years putting Python inside museum walls. Now I'm putting AI agents inside sandboxes.

Source: Dev.to

Introduction

I’ve spent over a dozen years experimenting with Python in environments where it traditionally doesn’t belong. From mobile app tooling to interactive museum installations, the common thread has always been making Python work against the odds.

Python in Unlikely Places

Kivy and Mobile Tooling

I co‑created Kivy, the cross‑platform UI framework, and then built the surrounding ecosystem that lets Python run on phones:

• pyjnius – call Java from Python
• python‑for‑android – package Python apps for Android
• kivy‑ios – bring Kivy to iOS
• buildozer – automate the entire build chain

If you’ve ever shipped a Python app on a mobile device, you’ve likely used one of these tools.

Museum Installations

For more than 12 years I freelanced on interactive installations for museums and cultural institutions across France:

• Touch‑screen exhibits at the Palais des Beaux‑Arts in Lille
• Digital projects for the Bibliothèque nationale de France
• RFID‑based installations at the Philharmonie de Paris

These projects shared three constraints:

1. Python was the core language.
2. Touch screens were the primary interface.
3. Cursed deployment scenarios required the software to run invisibly in environments not designed for it, often under the watchful eyes of officials during pre‑opening visits.

Music and Coding

I also make music (piano and modular synthesis) and view coding as a similar creative process: you layer elements, tweak them until they feel right, and express something that’s hard to articulate otherwise. I’m more comfortable with a terminal than with my voice, which is why I’m writing this post instead of recording a podcast.

AI Agent Security

These days I’m focused on AI agent security. After becoming comfortable with AI‑assisted coding, I realized I had zero visibility into what my coding agent was doing on the network. To address this, I built:

• greywall – a tool for monitoring network activity of AI agents
• greyproxy – a proxy that gives you insight into the agent’s external communications

I’ll be writing about what I discovered, what I built, and the lessons learned along the way.

Closing

If you care about open source, security, or building tools that reveal what’s really happening under the hood, I think we’ll get along. Happy to be here!