Home / Workshop, capstone and exam

Warm-up, assembling a prompt and ticking your checklist

Rebuilding a complete prompt from memory and ticking your own practitioner's checklist before any project reveals the reflexes still missing, where rereading an example would only test reading.

The first project in this module does not write a single line of code: it checks whether the reflexes built up over the course come back on their own, with no notes open alongside. The workshop asks you to rebuild a complete prompt from memory, then tick a practitioner's checklist before moving to the next project. A missing reflex here costs a minute to spot; the same gap discovered in the middle of the capstone costs an entire session started over from scratch.

What a prompt rebuilt from memory reveals

Writing a prompt with an example in front of you, and rebuilding one from memory with nothing to consult, do not draw on the same skill. The first tests reading, the second tests what remains once the reference is taken away. A forgotten role and context, a constraint left unstated, an output format left implicit: each missing block in the rebuilt version points to a reflex not yet installed, rather than a detail without consequence.

Brique 1, rôle et contexte : qui répond, et à partir de quelles informations
Brique 2, tâche précise : ce qui doit être produit, en une phrase
Brique 3, format de sortie attendu : la forme exacte du résultat
Brique 4, contrainte à respecter : ce que le résultat doit respecter ou éviter
Brique 5, exemple concret : un cas déjà résolu qui montre le format visé

Assemble these five blocks, without looking at an existing example, for a writing task, a coding task and a research task. The block that goes missing most often from one task to the next shows where to focus attention before tackling the capstone. A block present in all three versions, on the other hand, no longer deserves the time usually spent on it: it is already acquired.

The checklist closes the loop

Once the three prompts are assembled, the practitioner's checklist gets ticked point by point, not all at once during a final reread. Each box corresponds to a concrete action practised earlier in the course: checking a source before citing it, testing an automated check with a case it must catch, reading a command's output before asserting a result. A box ticked without being able to name the precise action that justifies it does not count; it gets unticked and redone.

The same principle applies to the three prompts themselves: a rebuilt prompt that resembles the original without being able to explain why each block is there is worth no more than a box ticked at random. Understanding is proven by naming the role of each block, not by producing text that sounds right.

This double check, rebuilt prompt and ticked checklist, prepares directly for choosing the end of course project, which is made on what these two exercises have just revealed rather than on an arbitrary preference.