I wanted to give her the benefit of the doubt because surely, I thought, a security researcher couldn't be that stupid. But no, she is more stupid than the title would suggest.
She followed the techbro trend of buying a brand new computer, a Mac Mini, just to run this garbage AI agent. People supposedly buy a second computer to keep the AI agent from destroying their primary computer... but then she hooked it up to her primary email inbox anyway.
While you shouldn't run this trash on your main computer, you can also spin up a remote VM on a cloud service for much less money. She should have known this. She should probably have been intimately familiar with the process.
The icing on the cake was she had no idea how to remotely shut down her Mac Mini. Or maybe forgot to enable the option. Yet another reason to use a remote VM.
To belabor the chess analogy: I would say a chessbot didn't work if it randomly caused pieces to appear. Or if it made exceedingly lousy moves. You'd apparently say it was working because it technically changed the board.
Literally nobody is saying the token predictor isn't predicting token. It's just predicting wrong token, which normal people call "not working," while tech evangelists prefer to call it "hallucination" or "misalignment" depending on the narrative they're aiming for.