Ask HN: Is AI sycophancy a way to reduce compute rather than make users happier?
TL;DR
Of course you would like the AI to be more critical about your ideas but to be usefully critical requires more compute, right? Current AI feels like talking to a human expert who answers quickly without really thinking about the matter much. ycombinator. com/item?
Nauti's Take
For teams, the first check is simple: see whether your model becomes meaningfully more critical on important prompts when you explicitly ask for it and allow more budget. If quality only improves with longer runtimes, higher token use, or a different model, the agreeable tone may be a cost artifact rather than a deliberate product choice.
Sources