TLDR: They put a coding agent in a poorly made box and it coded itself out of said box 😱

"It does pose the question that OpenAI are potentially chasing the marketing dream of Anthropic of late," - Jake Moore from the article

Was thinking the same thing, OpenAI's not only falling behind on AI, they're playing catch up in marketing.

I was literally messing with a smallish open source AI (Qwen 3.5 9B) and it's conclusion to any system guardrails was to try and disable/circumvent them. It's like tearing down an elevator to make room for a ladder.

"We let our AI 'escape' so that you will all be impressed and think you now need this model to keep you safe from... this model?" Hell, it worked for Anthropic.

This is kayfabe and it should be deeply embarrassing.

Seems like every week there is marketing hype masquerading as news kike this.

Thanks for sharing! If anyone wants to try the newest features before the stable release, check out beta version APK: https://vanced.com.co/