Prompting is requirements gathering with worse vocabulary
The teams that get good results from AI aren't better at talking to models. They're better at saying what they want.
By Nzubechukwu Cyprian · Pharmacy student at UNN, full-stack developer
Every failed prompt I've written was really a failed specification. That took me embarrassingly long to admit.
Consider the difference. "Summarise this document" is a wish. "Summarise this incident report in five bullets: what happened, what broke, who acted, what the impact was, what the follow-up is. If any of these are missing from the document, say so rather than guessing" — that's a spec. The model's quality jumped not because I found magic words, but because I finally decided what I actually wanted.
What spec-writing looks like in a prompt
- define the output shape before the task ("return JSON with keys X and Y")
- state the audience ("for a non-technical manager")
- state the failure behaviour ("if unsure, return unknown — do not guess")
- give one worked example; two if the task has edge cases
Why this matters more now
When AI writes the first draft, the leverage moved upstream. The person who can state precisely what should exist — inputs, outputs, edge cases, failure modes — gets dramatically more out of the machine than the person hunting for secret phrases.
Funny thing is, this was always the skill. The best engineers I know were already writing tickets like prompts before prompts existed.
One email when a post lands
No schedule, no spam. Just the posts worth reading — new writing on software, AI, products, and the Pharmacy × Tech life.
Prefer a feed reader? Grab the RSS feed.
Keep reading
AIThe model is not the product
Everyone debates which model is smartest. Almost nobody talks about the ninety other things standing between a prompt and a shipped feature.
AIMy AI workflow is boring, and that's the point
No autonomous agents rewriting my codebase. No prompt wizards. Just a disciplined loop between a draft, a review, and a test suite.
AII gave an AI agent shell access. Here's what I learned.
A week of running an autonomous coding agent on a real project taught me where the ceiling is — and it isn't intelligence.