The AI Learning Hub Journal

When It Is Not What You Pictured

The first version will not match the picture in your headyour head holds a hundred details you never wrote down — the result reflects only the part you saidIT IS A DRAFT TO REACT TO — NOT A VERDICT ON YOUR IDEA, OR ON YOUskilled builders are not better at first prompts — they are quick to say precisely what is wrongWHAT CAME BACKWRONG UNDERSTANDINGit built something elseentirely — your descriptionmissed the pointrestate the goal fromthe top — do not tweakWRONG DETAILSthe right thing with wrongspecifics — colour, order,a missing fieldname the exact fix,one at a timeWRONG BEHAVIOURlooks right, but does thewrong thing when used —nothing on screen warns youtake this pile mostseriously — it hidesTHE CLASSIC TIME SINK — ENDLESS SMALL TWEAKS TO A MISUNDERSTANDINGwhen the understanding is wrong, adjusting details cannot fix it — go back and restate what you wantSOMETIMES WHAT CAME BACK IS BETTER THAN WHAT YOU ASKED FORseeing a thing exist reveals what thinking never does — update the spec rather than defending itSORT WHAT IS WRONG INTO THREE PILES BEFORE YOU REACTeach pile needs a different response — and mixing them up is where the time goes
A first result that differs from the picture in your head is a draft, not a failure — sort what is wrong into three piles before reacting.

This Is Normal, Not Failure

The first version of anything comes back different from the picture in your head. That is not the tool failing, and it is not you prompting badly. It is the ordinary result of a description being less complete than a mental image. Your head contains layout, tone, spacing, the order things happen, and a hundred details you never wrote down. What came back reflects only the part you managed to say. So the first result is not a verdict on your idea. It is a draft, and drafts exist to be reacted to. People who get good results are not writing perfect first prompts. They are unbothered by an imperfect first result, and quick to say precisely what is wrong with it.

  • A mental image always contains more than the description you wrote
  • The first result is a draft, not a verdict on the idea or on you
  • Skilled users are not better at first prompts; they react faster and more precisely
  • Expect two or three rounds on anything visual as the normal case

Separate the Three Kinds of Wrong

Before reacting, sort what is wrong into three piles, because they need different responses. Wrong understanding: it built something else entirely. Your description missed the point, so it needs rewriting rather than adjusting. Wrong details: the right thing with wrong specifics — wrong colour, wrong order, missing field. That is a small correction. Wrong behaviour: it looks right, but does the wrong thing when used. Take that pile most seriously, because it hides. Mixing these up wastes time. The most common version is making repeated small tweaks to something that was built on a misunderstanding and needed restating from the top.

  • Wrong understanding: restate the goal instead of adjusting details
  • Wrong details: name the specific fix, one at a time
  • Wrong behaviour: looks fine, acts wrong — the most important pile
  • Tweaking a misunderstanding forever is the most common time sink

Sometimes You Were Wrong

Occasionally what comes back is better than what you asked for. Or it reveals that your idea did not quite make sense. A screen you imagined turns out to need a step you had not thought of. A feature you were sure about looks pointless once it exists. This is one of the genuine gifts of building fast. You find out cheaply, in an afternoon, what used to take a quarter and a budget to discover. Notice when your reaction is "that is not what I said" versus "oh, that is actually better". Be willing to update the spec rather than defending it. The document is a tool for thinking, not a contract you signed with yourself.

  • Seeing a thing exist reveals problems that thinking about it never does
  • Update the spec when reality teaches you something; do not defend it
  • Cheap building means cheap learning — that is most of the value
  • Distinguish "not what I asked for" from "better than what I asked for"

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.