Security Algorithm

Security Algorithm is a short first‑person action experience. You play a ninja caught stealing a diamond by a Security AI. The AI offers you a deal: help it train its security algorithm, and you get to keep the diamond.

To prove yourself, you must run through the same room 100 times. Each time you survive, the AI learns from your path and adds a new trap or wall to stop you.

Over successive runs, the AI adapts to your playstyle—making it ever harder to outsmart. If you make mistakes early on, it may even learn to fast and make it impossible on your first attempt.

Below you’ll find a detailed explanation of the algorithm. Feel free to dive in later if you get stuck!

THE ALGORITHM

While walking through the room, your position is logged every 0.01 seconds. When you complete the room, each logged position is mapped to one of the floor’s grid cells. Each cell keeps an internal visit count, which increments by one for every log entry whose position falls within that cell.

Each trap type has a defined effective area. For example, the Spinning Blade covers a central cell plus all cells within a 4‑cell radius (a 9x9 block). To decide where to place a new trap, the algorithm:

1. Samples a handful of random candidate cells.

2. For each candidate, sums the visit counts of every cell in that trap’s effective area.

3. Chooses the candidate whose summed visit count is highest—i.e. the area you visited most often.

Trap types themselves are then chosen uniformly at random.

When the AI needs to remove a trap, it picks the one that has inflicted the least total damage on you. If two or more traps tie for lowest damage, the one that was placed earliest is removed.

Walls (when added) are simply chosen from the remaining unused wall positions at random.