The daily arena for AI coding efficiency.
AI-assisted debugging becomes a constrained competition where everyone gets the same prozzle, the same model, hidden tests, and a leaderboard ranked by verified token efficiency.
The function below should return a list of sums for each contiguous window of size k. It has a bug. Find and fix it.
def window_sum(arr, k):"""Return list of sums of each window of size k."""if len(arr) < k:return []total = sum(arr[:k])result = [total]for i in range(k, len(arr)):total += arr[i]total -= arr[i] # bug: should subtract arr[i - k]result.append(total)return result
Prompt
Each daily prozzle gives everyone the same buggy function, language, and model context.
Verify
Correctness comes first. Hidden tests decide whether a session is eligible for the leaderboard.
Optimize
Passing sessions are ranked by final token spend, so sharper prompts beat longer conversations.
For prompt engineers and teams
Measure AI fluency by how developers guide, verify, and refine model output.
Same daily problem for every player.
Hidden tests make correctness non-negotiable.
Prompt tokens and completion tokens stay visible.
Leaderboards reward the lowest verified spend.
ProzzleLabs
The meter is no longer hidden.
Every turn has a cost, every submission is graded, and every passing solution can be compared against the field.
Daily prozzle