Turn any prompt you've used twice into a template

You wrote the perfect prompt three weeks ago. It nailed the tone, the structure, the format. You will never find it again. It's buried in a chat you can't search, and right now you're rebuilding it from memory, worse than the original.

This happens to everyone who uses AI seriously, and almost nobody fixes it. The reason is subtle: a good prompt is an asset, but it doesn't feel like one. It feels like something you typed. So you don't keep it the way you'd keep a spreadsheet or a doc. You keep it nowhere. And then you rewrite it, from scratch, every time.

There's a simple rule that ends this: any prompt you've used twice should be a template, not a one-off.

The difference between a prompt and a template

A prompt is a thing you typed once for one situation. A template is the same thinking with the situation-specific parts pulled out into blanks you fill in.

Take a prompt you probably write some version of constantly:

Write a cold email to a VP of Marketing at a mid-size SaaS company, offering a free audit of their content workflow. Keep it under 90 words, casual but specific, one clear ask.

Used once, that's a prompt. But you write this shape all the time — different role, different company, different offer. So turn the specifics into blanks:

Write a cold email to a $role at a $company_type, offering $offer. Keep it under $word_limit words, $tone, one clear ask.

Now you're not rewriting the structure every time. You're filling in four blanks and keeping the part that actually took you three iterations to get right: the framing, the constraints, the "one clear ask." That framing is the asset. The blanks are just today's details.

Why this compounds and rewriting doesn't

When you rewrite from memory, you start over at zero every time, and you usually forget the good bits. The 90-word limit that made it punchy. The "casual but specific" that stopped it sounding like a robot. Each rewrite is a little worse and a little slower.

When you templatize, the opposite happens. Every time you use the template and it falls short, you improve the template, not just today's output. The word limit gets tuned. You add "no exclamation points." You add an example of a good one. Six months in, that template is the best cold-email prompt you've ever had, because it absorbed every lesson from every send.

That's the whole game with AI at work: the people who get better are the ones whose prompts compound. The people who stay the same are rewriting from scratch every day and calling it prompting.

How to start in the next ten minutes

You don't need a tool for the first step. You need a habit:

• Notice the second use. The moment you think "I've written something like this before," stop. That's a template waiting to happen.
• Pull out the blanks. Look at your prompt and mark every part that changes case to case. Those become your variables. Everything else is the reusable structure.
• Keep them somewhere you'll actually reopen. This is where most people fail. A Notes app you never open is not a library. It has to be one search away, or you'll retype from memory anyway.
• Improve the template, not the output. When a result disappoints, fix the template so the next hundred uses inherit the fix.

The last two are where it gets hard by hand. Keeping prompts findable, versioning them as you improve them, and filling in the blanks without hunting for the file — that friction is exactly why most people's "prompt library" is a graveyard of screenshots.

That's the problem Korvai solves: a real prompt library with $name variables, so you save the structure once, fill in the blanks in a click, and every prompt you've earned is one search away instead of lost in a chat. But the rule works with or without a tool. Start noticing the second use. Start pulling out the blanks. Stop rewriting your best work from memory.