Article · 2026-07-14 · 6 moments

simonw/pedalican

simonw/pedalican Clearly I wasn't paying attention when these were first announced back in May, but today I accidentally activated a "pet" in Codex Desktop - a little animated robot, reminiscent of Clippy - and then learned you can create your own. So I did, and now I have a cute little pelican on a bicycle bouncing around my desktop giving me updates on my Codex tasks. Your browser does not support HTML5 video. The most interesting thing about this process was watching how the c ✦ AI generated

Read original ↗

HOST Simon Willison

Example◆

A single gpt-image-2 output image, produced from a highly structured prompt specifying character identity and a flat magenta chroma-key background, was compiled into the waving.gif animation.

The waving.gif animation was compiled from one gpt-image-2 generated image, created via a prompt with a defined structure: pet identity description plus instructions for a centered pose on a pure magenta background sized for animation.

transcript

Simon Willison: Create one clean full-body reference sprite for Codex pet Pedalican. Pet identity: A compact adorable baby pelican with a round cream-white body, soft coral-orange bill and feet, riding a tiny sky-blue bicycle [...] Place a single centered pose on a perfectly flat pure magenta #FF00FF chroma-key background. Keep the full pet visible, compact, readable at 192x208, and easy to animate.

Anecdote

GPT-5.6 Sol itself invented the name "Pedalican" for the custom pelican-on-a-bicycle pet.

Willison confirms that the model, not he, came up with the pun name "Pedalican" for the pelican-on-a-bicycle pet, and says he likes it.

transcript

Simon Willison: And yes, GPT-5.6 Sol did come up with the name "Pedalican". I like it!

Anecdote◆

GPT-5.6 Sol operating in xhigh mode autonomously designed the custom Codex pet, using multiple rounds of gpt-image-2 to generate the sprite assets.

Simon Willison describes asking Codex Desktop for a custom pelican-on-a-bicycle pet and having GPT-5.6 Sol xhigh handle the entire creation process, iteratively calling gpt-image-2 to produce the needed sprite images.

transcript

Simon Willison: The most interesting thing about this process was watching how the custom pet was created. I told it I wanted a custom pet that was a pelican riding a bicycle and GPT-5.6 Sol xhigh did the rest of the work, using several rounds with gpt-image-2 to generate the necessary sprite assets.

explains mechanism · 1gives example · 1

Fact

The full generation process, including every intermediate image, combined sprite sheets, and animation GIFs, was documented and published in a GitHub repo.

Willison had the model keep extensive notes and record every intermediary step, publishing all generated images, sprite sheets, and animation GIFs (like waving.gif) in his repo.

transcript

Simon Willison: I had it make extensive notes and record all of the intermediary steps. My GItHub repo includes every generated image and combined sprite sheet, plus GIFs for each of the animation loops such as this one, called waving.gif

gives example · 1

Fact

The key implementation details behind Codex pet creation are open source, specifically the hatch-pet skill (openai/skills) and the imagegen skill (openai/codex), both under the Apache 2.0 license.

Willison notes that the two skills powering the pet-creation feature, hatch-pet and imagegen, are open source under Apache 2.0, letting him dig into how the mechanism actually works.

transcript

Simon Willison: The key implementation details are open source - these two skills in particular, both Apache 2.0 licensed: hatch-pet from openai/skills, imagegen from openai/codex

Context

This project serves as a useful real-world example of using image generation to create simple, game-ready sprites.

Willison frames his exploration of the Pedalican pet-creation process as motivated by his interest in finding examples of image generation being used to produce game-ready sprites.

transcript

Simon Willison: I've been looking out for example of ways to use image generation to create simple game-ready sprites, so I spent some time digging into this mechanism to see how it works.

gives example · 2

Highlight slides

Codex Desktop Builds a Custom Pet Autonomously✦ from: GPT-5.6 Sol operating in xhigh mode autonomously designed the custom Codex pet, using multiple rounds of gpt-image-2 to generate the sprite assets.One Image Became waving.gif✦ from: A single gpt-image-2 output image, produced from a highly structured prompt specifying character identity and a flat magenta chroma-key background, was compiled into the waving.gif animation.Sprite Generation via gpt-image-2✦ from: GPT-5.6 Sol operating in xhigh mode autonomously designed the custom Codex pet, using multiple rounds of gpt-image-2 to generate the sprite assets.Prompt Structure Behind the Sprite✦ from: A single gpt-image-2 output image, produced from a highly structured prompt specifying character identity and a flat magenta chroma-key background, was compiled into the waving.gif animation.

Related episodes

Audio · The Changelog

Community perspectives on Elastic vs AWS (Interview)

→

Article · Interconnects

Teaching Everyone to Fish for Tokens

→

Video · All-In Podcast

OpenAI vs Anthropic IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts

→