[blastonico]

Oh boy, this really seems to be hallucination.

Guys, chatbots are mostly token generators, they don't run programs e give you responses...it's not a simple shell program, it computes things in GPU and return tokens, in which are translated back to English.

[MattGaiser]

This is very out of date. They now often trigger tooling and return the outputs of the tooling.

[blastonico]

Not really. You're referring to agents, but the model doesn't always require agents, and the public chatbot is not connected to a shell freely evaluating arbitrary commands.

[brookst]

This is false.

Modern chatbots use tool orchestration where the LLM is given a list of tools it can use. Tools include things like searching the web and, yes, executing code. The LLM can generate arbitrary code (typically Python) and call a tool to execute it and return results. The tool takes the Python input, launches it (in a container in this example) and returns the output, which the LLM can further process.

[blincoln]

> and the public chatbot is not connected to a shell freely evaluating arbitrary commands.

[ citation needed ]