Skip to content
Nuvelorb

Tested, not guessedTest plan

AI coding help for a non-coder building a small app

Task: Someone who has never coded builds a small working app by following one AI tool's instructions.

Demos make app-building look instant. We sit a real non-coder in front of each tool with the same small app idea and count what it takes to get something that works.

This is a test plan. Nothing has been run yet, so there are no scores or verdicts below.

How we tested

  1. One written app brief and an acceptance checklist (what "working" means), the same for every tool.
  2. A tester with no coding experience uses one tool at a time, starting fresh, and only does what the tool tells them.
  3. We stop at a working app or after 90 minutes.
  4. Recorded per tool: did it work, number of messages to get there, how many times the tester was stuck, every checklist item that passes, cost, and total time.
  5. A person also scores how clear the explanations were.

What we scored

Checklist items passing · human-judged · weight 3
Out of the acceptance checklist (enter the number passing).
Messages to working · human-judged · weight 1
How many messages the tester sent. Lower is better (cap 50).
Times stuck · human-judged · weight 1
Times the tester couldn't follow an instruction. Lower is better (cap 10).
Explanations a beginner understands · human-judged · weight 1
Were the steps explained in plain words?

Inputs every tool received

  • App brief and checklistPlaceholder

    Placeholder: brief and checklist to be provided.

Inputs still needed

  • The app brief (for example: "a one-page app to log my shop's daily expenses by category and show the monthly total, saved on my computer").
  • The acceptance checklist: 5–8 things the app must do.
  • A tester who has never written code, and the tools and plans to test.