Lore
Stated Confusion as a Prompting and Review Technique
Overview
Two related moves treat a reviewer's or user's genuine lack of understanding as a legitimate, actionable input to an agent, rather than something to hide until full expertise is reached:
- "I don't get this" as code review: when reviewing agent-generated code you don't fully understand, respond with your actual confusion or suspicion ("I've done my best to investigate and I'm pretty sure this is wrong") as the review comment itself. This prompts the agent to explain or fix the code, or the comment can be escalated to a teammate — it does not require the reviewer to first reverse-engineer full understanding.
- "Explain it like I'm five" prompting: when working outside your expertise (e.g., debugging CI internals), explicitly tell the agent you don't understand and ask it to explain its reasoning or a failure in maximally plain, non-expert language before you accept its fix.
Both treat honestly-expressed non-expertise as more productive than either blocking on expertise-building first or blindly accepting agent output.