2 min lesson
Know the developer's world, not just the UI
Match each case in "Know the developer's world, not just the UI" to the signal and response that fit it.
Step 1 of 2
Know the developer's world, not just the UI
Trust in AI output: when do they accept, verify or reject a generation and what breaks the trust?
Flow state: the cost of a context switch is high, so friction that breaks flow is felt sharply.
Review burden: AI shifts work from writing to reviewing and reviewing bad output can cost more than writing from scratch.
Model choice: which model for which task and how users reason about the tradeoff at all.
Lead with a friction, then turn it into a research question with a hypothesis. “The moment Agent makes a change I can't fully see, I hesitate before accepting - and I'd bet that hesitation is the real adoption barrier, not feature gaps. I'd test it with a study pairing session replays against self-reported trust.” A papercut that becomes a study reads as a daily user and a researcher at once.
Do not fake a take you do not hold. This user base smells it and so does the team. If you have not used a surface deeply, say what you observed and frame the rest as a hypothesis you want to test, not a fact you assert. Honest curiosity grounded in some real use survives; a confident wrong claim about the product ends the read.