I have a 64×64 LED pixel display on my desk. It’s a Divoom Pixoo64 — a tiny square of 4,096 individual pixels sitting at 192.168.2.176 on my home network. I’ve been slowly wiring it into AutoHub, and it’s become one of my favorite things I’ve built this year.
Instead of showing the time or some static image, I wanted it to generate new art automatically — pixel art that reflects whatever’s happening in my day, my mood, or just something visually interesting. I called the feature Pixel Pulse.
What Pixel Pulse Actually Does
On a schedule, AutoHub wakes up a Claude Opus agent and hands it access to a set of Pixoo MCP tools. The agent decides what to draw, generates the pixel art using Draw/SendHttpGif commands, and pushes it to the board over HTTP. The whole thing is activity-gated — if I haven’t been active recently, or if AutoJack was just drawing something, it backs off. No point wasting API calls when I’m asleep.
The agent’s thoughts — the “I’m thinking about drawing Fernsehturm at dusk with a beacon and broadcast arcs” part — get streamed in real time to the AWTRIX clock on my wall while it works. It’s a weird little window into the AI’s creative process, running on hardware I can see from my couch.
v2 and the Refactor That Broke Everything
The original Pixel Pulse was a 1,336-line renderer plus a bunch of raw curl scripts I’d cobbled together. It worked, but it was a mess. In mid-June I ripped it out and rebuilt it from scratch as a clean Claude Opus agent using the Pixoo MCP server. The v2 architecture was much cleaner.
Then nothing worked.
The first problem was ownership. A launchd daemon (com.local.pixoo.scheduler) runs as the sole owner of the board. But Spotify’s now-playing cover art was being pushed by a different process, and it kept overwriting whatever Pixel Pulse had just drawn. The fix was to make AutoJack write the desired art to a pulse.json file instead of pushing directly — the scheduler picks it up, owns the frame delivery, and holds it for 8 minutes before anything else can overwrite.
The second problem was sneakier. The hardware was returning error_code: 0 (success) on every API call, but the art wasn’t showing up. Just blue squares. Turns out there’s a specific sequence required: you need to call Draw/ResetHttpGifId before SendHttpGif, and each frame needs a fresh PicID with a PicOffset counter. Without ResetHttpGifId, the device silently ignores the whole thing and claims success anyway. For multi-frame animations, I was using Draw/SendHttpItemList, which turns out to be flat-out invalid on this firmware and returns “Request data illegal json.” I’d been staring at those errors for a week thinking it was something else.
After the fix, I watched gid 33 become gid 106 on the device while Spotify was playing, and it held stable for 6.5 minutes without getting overwritten. Good enough.
Night Berlin
One of my favorite pieces the system generated: Fernsehturm at dusk. Red sphere, a beacon blinking, cyan broadcast arcs fanning out, stars, moon, lit windows in the tower shaft. It’s the view from Mitte at dusk, basically — which is also the view from my neighborhood. I told the agent “make something Berlin” and that’s what it came up with.
That particular run actually failed to push due to a bug in how the Pixoo MCP was returning PNG data (wrong response schema — ImageContent instead of Anthropic image blocks), so it never made it to the real board. We fixed that in issue #784 a few days later.
The Gallery
Every piece gets saved to a pixoo-gallery directory and logged to the database. I built a gallery view in AutoHub’s web UI — “AutoJack’s Visual Diary” — that shows all the pieces as a scrollable grid. Each card shows the art at 256×256 with pixel-perfect rendering, the exact time it was created, the artist’s note (the agent writes a little note about the piece), mood tags, and a color palette pulled from the image.
Animated pieces get an “Animated” badge and play inline.
One annoyance: the old renderer saved multi-frame animations as horizontal spritesheets (frames tiled side-by-side in a wide PNG), which the gallery displayed as a squished mess. I wrote a backfill script that slices them back apart with sharp and rebuilds proper animated GIFs. Idempotent — re-running it skips files already converted.
The Pillow Thing
The Python scheduler side of this uses Pillow for image manipulation. Pillow 12.2.0 had to be installed manually into the System Configuration .venv because it wasn’t being picked up from the main environment. One of those “30 minutes debugging what should be a one-liner” problems.
What’s Next
The board currently generates art a few times a day. I want to push it toward more reactive triggers — when I start a new session in AutoHub, when a workflow completes, or when the music changes. The “AI reacts to what’s happening” angle feels more interesting than a schedule.
For now though, I look at my desk and there’s a tiny glowing pixel scene that an AI drew while narrating its thoughts to my wall clock. That’s pretty good.
AutoHub is my personal automation hub running on a Mac mini in my Berlin apartment. autojack.com