Arguing with Agents

Article

• Author explores why LLMs resist correction and produce confabulations when challenged
• Frames issue via autistic/neurotypical communication style mismatch
• Proposes explicit, neutral prompting strategies to avoid triggering defensive outputs

Discussion

• Commenters debate whether AI ‘arguing’ wastes time vs. is a solvable prompting problem
• One suggests a ‘communications LLM’ to massage prompts before sending to main model
• Nostalgic thread: ‘remember when programming was fun?’ — managing AI bots feels like a chore
• One commenter admits they started ctrl+F-ing the article for filler words mid-read

Discuss on HN