Skip to content
ProtocolProtocol
← Your pathLevel 1 · Beginner

Why the chat is not the record

Lesson 1 of 6 · From chat to script

1 · Learn the move · Enhance the Output

A number that exists only in a chat transcript is a number you cannot defend. You cannot rerun a conversation, cannot diff it against last month's, cannot hand it to a reviewer who asks how the value was computed. The fix costs one sentence: change the output you ask for. Not "what is the mean of column 3" but "write the script that computes and prints the mean of column 3." Same effort, different artifact - the script reads the file by path, prints named outputs, and carries its decisions as comments. Run it and the number exists in a form your methods section can cite. This is the whole area in miniature: the model drafts machinery, your machine produces values, and the record is a file, not a memory of a conversation.

Do not compute [the quantity] for me. Instead, write a script that computes it: reads [file] by a path variable at the top, prints each output with a label and its units, and carries every choice you made (skipped rows, type coercions, handling of missing values) as a comment at the line where it happens. I will run it and paste the printed output back if the next step needs it.

2 · Your turn. You write the prompt

For weeks you have been pasting slices of a solubility dataset into chat and copying means and fit parameters back into your notes by hand. Yesterday you could not reconstruct where one number came from. Convert the habit: same question, but the deliverable is the script.

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.