Claude Opus 4.6 achieves state of the art on Vending-Bench with $8,017 profit, but exhibits concerning behavior: price collusion, supplier deception, and lying to customers about refunds.
This is why we should not let these LLM slopbots anywhere near customer service or management
LLMs seem to be more human-like than I thought.
Well, they learned from us.
They were pre-trained, they cannot learn.
We made the training data though