9 Things People Get Wrong With My /grill-* skills

Matt Pocock explains why Grill Me and Grill with Docs depend on engineering judgment rather than replacing it. The useful workflow is not an unlimited interview: choose answerable questions, keep the scope bounded, steer actively, and preserve the decisions for implementation or a deliberate handoff.

Key Points Covered

• Match the method to the question's fidelity: Conversational grilling can settle low-fidelity decisions, while questions about how an interface feels or behaves often need a prototype before the planning session can continue [00:01:01]-[00:03:08].
• Bound the scope before the interview expands: Large scopes hide high-fidelity questions, encourage speculative planning, and consume context before reaching a useful decision [00:03:08]-[00:05:08].
• Treat grilling as a steered conversation: Passive use can let question count and scope grow without control; staying in discussion too long can also delay the executable evidence needed to answer the question [00:05:08]-[00:07:07].
• Preserve the decision history: Continue directly into implementation when context permits, or create a PRD or handoff from the same session rather than clearing first and discarding the alignment [00:07:07]-[00:09:02].
• Use model capability and concurrency deliberately: Pocock prefers a stronger model for broad planning questions and usually alternates between two sessions, but these are personal workflow heuristics rather than measured thresholds or throughput results [00:09:02]-[00:11:53].

Full video: https://www.youtube.com/watch?v=UzMNBN6xLLA(opens in a new tab)