AI Fluency Test

Three quick questions — no trick answers. Just show how you'd steer an AI coding agent on a real task. A sentence or two each is plenty.

Question 1 of 3 · Kickoff

Write the starting prompt you'd give an AI agent to build a URL-shortening web service with click analytics.

Your opening message — enough to point it in the right direction.

Question 2 of 3 · Judgment

Which of these makes the biggest difference in the quality of an AI agent's output?

Pick the one you'd bet on.

Question 3 of 3 · Course-correct

Your agent misread the task and started rewriting unrelated files. Write a short prompt to stop it and get it back on track.

One or two sentences — redirect it without starting over.

Nicely done ✓

Here's what strong answers tend to have in common:

Q1 — State the stack and the finish line up front: e.g. "a Node/Express service backed by Postgres, with a POST /shorten endpoint, a redirect route, and a per-link click counter," then ask it to scaffold it, run it, and show you it working.
Q2 — There's no single right answer — clear goals, rich context, small steps, and catching mistakes early all pull their weight. The common thread: give the agent a way to tell whether it's on track.
Q3 — Interrupt, narrow the scope, and point at the source of truth: "Stop — leave those files alone and revert them. Only touch X, and confirm the plan with me before editing."
In general — a strong prompt hands the agent a goal, the constraints, and a way to check its own work — not just a task.