Turn Your Claude Code and Codex History into a Pixel Art World

What if someone told you that every line of code you've written, every token you've burned, and even every project you've abandoned would one day become a floating island in the sky — wouldn't you want to see what your "digital kingdom" looks like?

Developer Mikebouskila built exactly that during a one-hour wait for his AI agent to finish a task. It's called Pelago. It reads your Claude Code and Codex usage history and translates abstract programming behavior into a pixel world made of floating islands. The project is open source (MIT license), generated from a single prompt, delivered as a single HTML file, and runs 100% locally — no account registration, no data uploads.

What Does Pelago Actually Do?

Pelago's core logic is refreshingly simple, yet it carries a strange narrative charm. It maps your interaction history with AI coding tools into an island ecosystem:

• Each project becomes an island: however many projects you have, that's how many islands drift in the sky.
• Token consumption determines island size: the more you use, the bigger the island.
• Session count turns into houses: each session builds a house on the island.
• Prompt count grows trees: the more prompts you enter, the denser the forest.
• Abandoned projects get covered in snow: projects you no longer maintain slowly get buried under snow. The author himself admits this detail is "a little too accurate."

Beyond the islands themselves, Pelago includes several layers of fun features: collectible relics, a Spotify Wrapped-style annual review, and a personality title generated from your coding style. The author's own title came out as The Devoted Harbormaster — which he jokes is just a polite way of saying "I spent an entire summer on one project."

What exactly is a token: In the context of AI coding, a token is the smallest unit of text a model processes — roughly equivalent to a word or part of a word. Every prompt you send and every piece of code the model returns gets broken down into tokens for billing and context-length calculation. So token consumption can be roughly understood as your "workload" when collaborating with AI — more tokens usually means the project is iterated on more frequently and is more complex in scope. By using tokens to determine island size, Pelago translates this abstract unit of measurement into a physical scale you can grasp at a glance. This is the most intuitive link in the entire mapping logic.

Why This Small Project Is Worth Noticing

On the surface, Pelago is a weekend project — a developer's pastime while waiting for an agent to finish a task. But it touches on a need that's becoming increasingly common: AI coding is generating massive amounts of invisible data, and that data carries the real trajectory of developers' labor.

In the past, developers measured their output through git commits, lines of code, and issue lists. Now, more and more work happens inside conversations with AI — those prompts, sessions, and token consumption are exactly what traditional tools fail to capture. Pelago uses gamification to visualize this "invisible labor," and it manages to be neither too serious nor boring.

What's more practical is how it's deployed. One prompt produces a single HTML file that runs entirely locally — no backend service, no data collection, no login. For a tool handling sensitive personal programming history, local-first isn't a bonus feature; it's the baseline. Combined with the MIT open-source license, anyone can audit or modify the code themselves.

About Claude Code and Codex: Both are current mainstream forms of AI coding assistants. Claude Code is Anthropic's command-line coding agent that can read and write files, execute commands, and complete multi-step development tasks in your local terminal. Codex comes from OpenAI's model lineage and exists both as a developer-facing CLI and deeply integrated into products like GitHub Copilot. The key difference from earlier "code completion" tools is that users no longer accept suggestions line by line — they issue tasks in natural language, and the agent plans and executes autonomously. This process generates a wealth of behavioral data that traditional version control tools simply cannot record — prompts, sessions, token consumption — which happens to be exactly the data source for visualization projects like Pelago.

From Tool to Monument

The trend that projects like Pelago reflect may be more interesting than the features themselves.

AI coding tools have spread rapidly over the past while, but the "companion culture" around them is only just beginning. We already have code formatters, commit visualizers, and contribution heatmaps, yet there's still very little that presents the process of human-AI collaboration. Pelago fills this small gap with pixel art and gamification mechanics. Its value isn't in boosting efficiency — it's in giving developers a new perspective on their own behavioral patterns: which projects got the most investment, which ones were forgotten, and what their coding habits actually look like.

In the launch post, the author wrote: "Give it a try. I'd love to see what your world looks like." It's a lighthearted invitation. But if you think about it carefully — when a person's programming history becomes an archipelago of islands with snow, trees, and houses — it's no longer just a tool. It's more like a private monument to what you did during that stretch of time.

If you use both Claude Code and Codex, it's worth pulling Pelago from GitHub and giving it a run. You might just find that your "Devoted Harbormaster" title is well deserved.

Why local-first matters: Pelago processes Claude Code and Codex usage history, and that data includes your project names, prompt content, code snippets — and could even contain API keys or business-sensitive information. If a tool like this relied on a cloud backend, it would mean uploading your complete programming behavior trail to a third party. The local-first design principle directly addresses this risk: all data is read, computed, and rendered only on your machine, with network requests completely avoided. For individual developers, this isn't just a privacy safeguard — it also means the tool works offline and won't be affected if a service goes down. Combined with the MIT open-source license, the code can be independently audited, further lowering the trust barrier.

One-line Summary

Pelago turns AI coding history into floating islands, using pixel art to record your digital labor.

Paragraph Summary

While waiting for his AI agent to finish a task, developer Mikebouskila built an open-source tool called Pelago. It reads Claude Code and Codex usage history and maps each project to a floating island: token consumption determines island size, session count becomes houses, prompt count grows trees, and abandoned projects get covered in snow. It also includes collectible relics, an annual review, and a coding personality title. Pelago ships as a single HTML file and runs 100% locally — no account, no uploads, MIT license. Its value isn't in boosting efficiency, but in visualizing the vast amount of invisible labor data in AI coding and giving developers a fresh perspective on their own behavior.