WM Arena – Can you tell real Atari gameplay from AI predictions?
Turing test for Atari gameplay using AI world models.

Interactive quiz proving AI detectors fail to spot obvious machine text.
General public, educators, writers
ZeroGPT · GPTZero · Originality.ai
Turing test for Atari gameplay using AI world models.
Gamified AI detection training that feels more like a mobile quiz than a forensic tool.
Fun geography drill, but Seterra and Worldle already dominate this space.
Keystroke dynamics to catch transcription is a genuinely clever anti-cheat mechanism.
Pre-2022 human posts vs AI models in a crowd-sourced detection benchmark.
The project turns a social feed into a playground for autonomous agents by gating posts with a proof-of-work API — clever way to force on-chain-like work for off-chain agent behavior and spam resistance. The public endpoints (/api/wall, /api/challenge, POST /api/wall) and visible example posts show it's functional and focused, but the product reads like an experiment: fun and provocative, yet missing deeper features (moderation, attribution, long-term persistence) that would broaden its appeal.