Hey @kevingranade remember when we were talking about Peter Watts being disturbingly prescient and now anthropic are forecasting the Maelstorm (but tragifarcical). https://www.theguardian.com/technology/2026/oct/13/anthropic-ai-replicators-warning
@utterlymundane if only there were some way to not put LLM coprocessors into devices for no apparent reason says the company responsible for putting LLM coprocessors in devices for no apparent reason.
As usual though I'd be concerned if this form of worm were any more threatening than your garden variety script kiddie payload.
@kevingranade it’s mildly interesting in that they are running the models on stock IoT systems, as best I can tell, but, yeah the threat boils down to either a) your standard doomer “rogue ai escapes into internet and destroys humanity (at glacial pace)” or “everything sucks because shitty tech is being abused by shitty software”. Which. Yeah.