Friend-Shaped Objects: A Case Study on My Use of Large Language Models

A LessWrong essayist reflects on using large language models as emotional support during an isolating overseas work contract, describing custom instructions designed to push back against sycophancy rather than flatter. The piece turns darker when the author describes a suspected malware scare, recounting how Claude's inconsistent reassurances left them more anxious rather than less. As LessWrong tells it, the essay closes on the author refusing the chatbot's suggestion to seek out a real person for support that night.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton