In which @mitsuhiko, whoâs on a tear recently, defends and promotes the use of the word âclankerâ to replace âagentâ (in the LLM sense): https://lucumr.pocoo.org/2026/5/26/clankers/
In which @mitsuhiko, whoâs on a tear recently, defends and promotes the use of the word âclankerâ to replace âagentâ (in the LLM sense): https://lucumr.pocoo.org/2026/5/26/clankers/
Wow, one of my gen-z coworkers just used the term "chopped" so I had to go look that up. And in the list of gen-Z slang I was reading, "clanker" came up too.

A clank is, essentially, what we in the outside world would call a robot. Clanks vary widely in form, function, intelligence, and complexity, though there appear to be some limits. For example, true human-level artificial intelligence is rare, only known to have been produced by particularly gifted Sparks such as van Rijn and (working from an example) Tarvek Sturmvoraus. A non-canonical sketch by Professors Foglio suggests that Agatha Heterodyne's Dingbots are as small as autonomous clanks...
@timbray @mitsuhiko my objection to "clanker" is it's a person-shaped word. "the clanker deleted my database" is still personifying, and it evokes the idea "person-like entity that does not deserve respect".
I don't think we can avoid personifying language for machines; that battle was lost a long time ago. but we can choose words that are less person-shaped. "clanker" and "agent" and "bot" are person-shaped. "LLM" and "program" and "algorithm" are less person-shaped.
@timbray @mitsuhiko I mostly use âbotâ for similar reasons.
If you ever doubt these systems are mindless, just ask them why they disobeyed, screwed up, etc.
I was being lazy.
I misread it.
I was being sloppy.
No excuse. I shouldnât have done that.
None of these make any kind of sense for an LLM, even though they are all exactly what a human might say. (This doesnât prove anything, but it shines a light.)
Itâs bullshit all the way down.