simonw/pedalican

agentictool-use

What happened

Simon Willison documents his experience creating a custom animated desktop "pet" (Pedalican) for Codex Desktop. He utilized GPT-5.6 Sol xhigh to orchestrate multiple rounds of image generation via gpt-image-2 to create sprite assets, compile them into animation loops, and generate sprite sheets. The process relied on two open-source Apache 2.0 licensed skills from OpenAI: hatch-pet and imagegen (part of Codex), demonstrating how agents use specialized tools to perform complex, multi-step creative tasks.

Why it matters

Demonstrates the practical use of open-source agent "skills" and multi-step image generation pipelines to automate asset creation.

The take

While a desktop pet is a toy, the underlying workflow is highly instructive. It shows how OpenAI is structuring agent capabilities into reusable, open-source "skills" (hatch-pet, imagegen). This modular skill architecture is likely how future agent ecosystems will scale, allowing models to seamlessly chain complex tasks like prompt-to-sprite-sheet generation.

Don't read this site daily. Get it in your inbox.

The daily brief and Sunday deep dive — distilled, scored, and opinionated. For builders only.