site banner

Friday Fun Thread for July 3, 2026

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

1
Jump in the discussion.

No email address required.

Anthropic tried this: they tasked Claude with running a vending machine in their office twice (#1, #2). In #1 it operated at a net loss, in #2 it eventually managed a small profit, until it was introduced to employees outside Anthropic, then it went bankrupt (suspiciously, Anthropic doesn’t include that on their blog).

I predict a bigger issue is that today’s LLMs are too suggestible: if an LLM were “officially” put in charge of a real company with real money, it would be a human or group unofficially running it. LLMs are still getting jailbroken, even Fable was allegedly 24 days ago when Anthropic universally suspended access. I also believe LLMs also struggle with high-level open-ended tasks, evidenced by their creative writing and software architecture.

But as LLMs get smarter and less easy to jailbreak, I’m interested to see more an LLM-run experiments, and I do think they have the potential to be much better leaders than humans.