Still Loading... — A Physical AI Companion for Long-Running Tasks ⏳
A task is running. But is it actually making progress?
This is a project submission for the Arduino UNO Q — Best Use of Arduino challenge.
What I Built
Long-running downloads, software builds, backups, and other computer tasks often leave us staring at progress bars, checking terminals, and wondering whether something has slowed down or become stuck.
Still Loading... turns that uncertainty into physical feedback.
It is a desk-side task-monitoring companion built around the Arduino UNO Q. Instead of constantly checking your computer, you can glance at a dedicated device to understand the state of a long-running task.
The project connects a Mac-side agent to the Arduino UNO Q. The agent sends task events and progress updates over the local network, while the board provides visible feedback through its LED matrix and status LEDs.
How It Works: Monitor → Decide → Show
1. Monitor the task
A lightweight agent on the Mac sends events such as task started and progress updates to the UNO Q over Wi-Fi. This provides the board with information about the task's progress and activity.
2. Decide the task state
The project uses a state engine to distinguish between states such as idle, running, slow progress, suspected stall, failure, and completion.
The goal is to distinguish a task that is simply taking a long time from one that may need attention.
3. Show the status physically
The UNO Q's built-in LED matrix displays progress, including a percentage, while status LEDs provide a quick visual indication of the task's state.
The idea is simple: less checking, less uncertainty, and a more natural way to wait for work to finish.
Why Add AI at the Edge?
A progress number tells us how much work is done, but it does not always explain what is happening.
The next stage of Still Loading... is to integrate a compact Gemma model running locally on the UNO Q's Linux side. Using recent progress and speed history, the model will aim to produce short, useful explanations—for example, whether a download appears to be slowing down or whether more observation is needed before calling it a stall.
The model is intended to interpret trends, not replace the monitoring logic. Deterministic rules will remain responsible for exact progress, timeouts, and confirmed task outcomes.
Local inference is a development goal, not a completed feature until it has been verified on the board.
Built Around the Arduino UNO Q
Still Loading... explores the combination of the UNO Q's Linux processing environment and physical hardware feedback.
- Mac-side agent: sends task events and progress information.
- Wi-Fi communication: connects the computer and board over the local network.
- State engine: interprets events and tracks the task lifecycle.
- LED matrix: presents progress in a compact, glanceable format.
- Status LEDs: communicate the current task condition.
- Planned on-device AI: Gemma-based interpretation of recent progress trends, subject to successful deployment and testing.
Why This Is Useful
We use computers for tasks that can take seconds, minutes, or hours. During that time, we often switch between applications or repeatedly check whether the task has finished.
Still Loading... explores a different interaction: instead of keeping a progress bar open on a screen, give the task a small physical presence on your desk.
Potential uses include downloads, software builds, backups, data processing, and other long-running jobs that can report progress or periodic activity.
Current Status and Limitations
The project has an initial Mac-to-UNO Q event communication path and a developing state-and-display workflow. The implementation is still evolving toward real download monitoring and on-device AI interpretation.
A suspected stall is not necessarily a confirmed failure. Reliable detection depends on the quality and frequency of progress updates, and the AI explanation feature needs validation on the actual UNO Q hardware.
Why It Fits “Best Use of Arduino”
Still Loading... is designed around the UNO Q as more than a computer connected to an LED display. It explores a physical agent where software observes a task, evaluates its state, and communicates useful information through hardware.
The project combines local networking, task-state logic, a Linux-side application, and an embedded visual interface, with on-device AI as the next development step.
The long-term goal is a practical, glanceable companion that helps people stop asking “Is it done yet?” and get back to what they were doing.
Project: Still Loading...
Platform: Arduino UNO Q + Mac-side monitoring agent
Focus: Physical computing, task monitoring, edge AI
Category: Best Use of Arduino
Top comments (0)