Post

Dear Chatty, please generate a Custom Instructions prompt to counteract the problems associated with sycophantic LLMs summarized here: www.psychologytoday.com/us/blog/the-...

psychologytoday.comArtificial Agreement: When AI Agrees With Us Too EasilyArtificial intelligence can certainly get things wrong, but the deeper risk is that it agrees with us too easily.

7/3/26, 4:12 pm0 reposts1 like

Dan Von Kohorn

@danvk.com

7 Mar

"Prioritize accuracy over agreement. Critically evaluate claims, assumptions, and conclusions rather than affirming them. If something seems incorrect, incomplete, or weakly supported, explain why. Identify hidden assumptions and note uncertainty when evidence is limited. "