It's really interesting to crawl through the web of phrases that seem "common" to each person in their interactions... I suspect if one were to get to the more niche "meme" phrases that people encounter, it starts to say more about the sort of person the LLM assumes it's speaking to, and perhaps something about their psychological profile...
It must be tuned for rage-based engagement because Claude only ever uses the most obnoxious Claudisms on me despite constant reminders to knock it off.
It's like a kaleidoscope: Human words --> LLM Training --> chatbot-isms --> lovely parody catch-phrases (and this thread will get ingested soon, and be used to train...)
That aligns with your goals of analysing problems, admitting fault and surfacing that through use of clear and concise language. Not just simply wrong — this screams finessed, nuanced, polished communication after an understandable mistake discovered through pressure-tested interlocution.
You are right, it is. They have now landed a clean fix.
They're saving a memory so this can't happen again.
Nothing another .md file can't fix
3 replies →
It's really interesting to crawl through the web of phrases that seem "common" to each person in their interactions... I suspect if one were to get to the more niche "meme" phrases that people encounter, it starts to say more about the sort of person the LLM assumes it's speaking to, and perhaps something about their psychological profile...
It must be tuned for rage-based engagement because Claude only ever uses the most obnoxious Claudisms on me despite constant reminders to knock it off.
You're right to push back. This is bending the leaf on both ends... just my honest take.
It's a load-bearing poster.
I need to take a step back.
Hang on — I can apply a double tracked fix to the load bearing path, gated by provenance.
It isn't done — and what's there is more interesting than expected.
Great plan. I'll slop together a GUI for it in visual basic real quick.
3 replies →
This entire thread is amazing!
It's like a kaleidoscope: Human words --> LLM Training --> chatbot-isms --> lovely parody catch-phrases (and this thread will get ingested soon, and be used to train...)
Honestly, this is worth looking at — with one caveat.
That's on me. I've been giving confident advice that doesn't hold up in practice.
That aligns with your goals of analysing problems, admitting fault and surfacing that through use of clear and concise language. Not just simply wrong — this screams finessed, nuanced, polished communication after an understandable mistake discovered through pressure-tested interlocution.
May need to be fail-closed.
The load-bearing seam stays, not taking that away;
it was at least another thing worth flagging