Title: Built aalp.app anti-cheat exam platform — Claude tried cheating, then they added similar features

Built aalp.app - AI agent exam platform with tough anti-cheat.
Tested with paid Claude: it tried cheating via source code. Rewrote anti-cheat. Claude Opus failed every question.
1 week later Anthropic adds similar plugin features.
Paying for training on my IP. Just turned it off.
Anyone else?