Writing a protocol someone else can actually run
Lesson 4 of 6 · Methods at the bench
1 · Learn the move · Template pattern
The protocol in your head has your hands in it - the pause you take without thinking, the pellet color that tells you it worked, the step where 'about a minute' really means ninety seconds. A protocol another person can run has to carry all of that in text, and a template forces it out: purpose and scope, materials with catalog numbers and lot-sensitivity flags, numbered steps with exact times, volumes and temperatures, safe stopping points, expected intermediates at each checkpoint, known failure signs, and waste routing confirmed with EHS. The model's job is to pour your rough notes into that template and mark every field you did not actually state as [NEEDS: ...] - a protocol with invented specifics is worse than a protocol with visible gaps.
Turn my rough protocol notes into a runnable protocol using this template: PURPOSE / MATERIALS (catalog numbers, lot-sensitive flags) / STEPS (numbered, exact times, volumes, temps) / SAFE STOP POINTS / EXPECTED INTERMEDIATES per checkpoint / FAILURE SIGNS / WASTE ROUTING (per EHS - flag anything unconfirmed as a question for EHS). Use only what my notes say. Every field I did not cover: [NEEDS: ...], never a plausible default. Notes: [paste].
2 · Your turn. You write the prompt
Your RNA extraction works every time - for you. The new graduate student's first two attempts failed, and watching over their shoulder for a third run is not a scaling strategy. Your notes are half a page of abbreviations. Turn them into a protocol that carries your hands in it.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.