I am flatly uninterested in letting an LLM execute arbitrary code as my user account on my laptop and maybe even sudo to root are you people fucking nuts
@kevinriggle um what who is suggesting this??
@azonenberg Probably this:
https://fly.io/blog/youre-all-nuts/
Quoting:
People coding with LLMs today use agents. Agents get to poke around your codebase on their own. They author files directly. They run tools. They compile code, run tests, and iterate on the results. They also:
* pull in arbitrary code from the tree, or from other trees online, into their context windows,
* run standard Unix tools to navigate the tree and extract information,
* interact with Git,
* run existing tooling, like linters, formatters, and model checkers, and
* make essentially arbitrary tool calls (that you set up) through MCP.
End quote
CC: @kevinriggle
@ColinTheMathmo @azonenberg @kevinriggle You know, as a visual artist with a basic understanding of the argument reading this, it is kind of hilarious. The author makes some good points: it's plausible that tools have been made to force LLMs to do something useful to people who are already baseline competent and capable of error-checking, within the confines of a dogshit system that values output and does not care about your or anyone else's career.
That this happens to also mean an insecure, environmentally disastrous, deeply unethical on multiple levels, completely unsustainable even a decade into the future, utter dumpster fire of a system (I am probably missing some problems!) that relies on working people in poorer countries to death and will inevitably hasten the ever-rapidly encroaching destruction of either the entire system or the planet itself is not something the software industry or community has ever cared about before it came for *their* jobs.
And the dude is really open about having decided they're not a person who gives a shit about second-order consequences or making it easier for other people to make personally-suboptimal but ethical choices, so, hey.