The project is Dozenal, a spatial arithmetic puzzle. The core mechanic is weirdly addictive but potentially frustrating: you move tiles on a grid to hit a total of 12, but the operation (add, subtract, multiply) changes based on how many empty spaces are between the numbers.
For example:
**0 spaces:** Addition (5 → 4 = 9)**1 space:** Subtraction (5 → _ → 4 = 1)**2 spaces:** Multiplication (5 → _ → _ → 4 = 20)
From a development perspective, this is a prime example of a modern AI workflow. It wasn't "vibe-coded" (where you just pray the LLM gets it right), but rather a hybrid approach. The AI handled the heavy lifting of the logic and layout, while the humans steered the design and refined the puzzle sets.
That said, the "human" side of the feedback is where things get messy. The learning curve is steep, and users are struggling with "mixed" arithmetic (combining two operations in one move). It proves that while AI can accelerate the deployment of a feature, it can't solve the UX puzzle of whether a mechanic is intuitive to a stranger.
If you're into grid-based logic or want to see what a human-AI collab looks like in production, it's a decent case study.
https://dozenal.game
Next How to Use a 14-Day Decision Loop for Career Pivots →