The specificity lever

You have the shape of a prompt: where, what, and sometimes why. This lesson is about the one dial that moves quality more than anything else, and how to turn it up without writing an essay.

Watch it fail first

Start Claude in your project and ask for something genuinely vague:

make my page better

Look at what came back. It probably changed several things at once, and some of them are not what you would have chosen. "Better" could mean the colours, the wording, the layout, the spacing, or all four. Claude had to guess, and you now have to read the whole file to find out what it decided.

Press Esc Esc and restore the code. Now do it properly.

One reasonable interpretation

in index.html, change the background to a light cream colour and make the heading font larger

Two changes, both named, nothing to guess. You can check it in five seconds instead of reading everything.

That is the whole test: is there exactly one reasonable interpretation of what I just asked for?

A quick checklist:

• Did you name the file? Not "the page", not "the code".
• Did you name the actual change? Not "improve", "clean up", "make it nicer".
• If something new is being created, did you say where it goes?

Being specific is not the same as being long. If a detail does not change the outcome, leave it out. You are aiming for no ambiguity, not maximum words.

The words that always cost you

These sound like instructions but contain no information:

When you genuinely do not know

Sometimes you cannot be specific, because you do not know enough yet. That is normal, especially early. The move is to be specific about the process instead of the outcome:

read index.html and tell me three things you'd improve, and why — don't change anything yet

Now you get options instead of a surprise. Pick one, then ask for that one thing.

This pattern — look first, then act — is worth more than any prompt template. It costs you one extra message and saves you the "that is not what I meant" loop.

Try it now

Ask for a deliberately vague change and watch the result. Then rewind, and ask for the same thing specifically. Compare how long it takes you to check each one.

That difference, repeated fifty times a week, is the whole skill.

Next: keeping the conversation focused as it gets long.