The Prompt Box Is the New Blank Page

I use AI every day. You probably do too.

So why does writing a good prompt still feel harder than it should?

ChatGPT, Claude, Gemini — pick your poison. They save me real time across work and life. But every time I open a fresh chat, I run into the same small wall: a blinking cursor in an empty box, and the quiet realization that I’m about to retype something I’ve already typed a hundred times before.

It’s a new kind of writer’s block. Not the absence of ideas; the absence of structure.

The gap nobody talks about

The goal is usually clear in my head. I know what I want the model to do. I know roughly how I want it framed. But translating that into a prompt that’s reusable, well-scoped, and worth keeping takes more effort than it should.

Context is king, as the saying goes. And context is exactly what gets lost the moment a chat scrolls off the screen.

So I end up doing one of three things:

1. Rewriting the same prompt from scratch, slightly worse than last time.
2. Digging through old chats trying to find that one version that worked.
3. Pasting a “template” out of a Google Doc that I forgot to update.

None of those feel like the future. They feel like a tax I’m paying for using tools that are otherwise excellent.

What I actually wanted

A place to build a prompt once. Save it. Reuse it by filling in variables, the way you’d fill in a form.

Not a wrapper around a specific model. Not another chat interface. Not a lock-in. Just a clean layer that lives outside the chat window and feeds whichever LLM I’m pointed at that day.

The closest analogy is snippets in a code editor. You don’t rewrite a for loop every time you need one; you expand a snippet and fill in the parts that change. Prompts deserve the same treatment.

The MVP

I built it. It’s called SpeedPrompt.app.

It is, by design, small. You write a prompt, mark the parts that change as variables, save it, and pull it back up later by filling in the blanks. Copy, paste into whichever model you prefer, done. No accounts to wrangle, no proprietary chat history, no opinion about which AI you should be using.

It’s an MVP, which is to say: it does the one thing I needed it to do, and not much more yet. That’s intentional. I’d rather ship something honest and useful than something bloated and clever.

Why I’m sharing it

Most of what I build is for me first. A tool to fix a friction I keep tripping over. Sometimes it stays a tool of one. Sometimes other people have the same friction and the tool turns into something larger.

SpeedPrompt is in that second moment now, where I’m curious whether the friction I feel is widely shared. If you work in admin, marketing, ops, or any role where you’re prompting AI several times a day, I’d genuinely like your take. What feels obvious? What’s missing? What’s wrong?

You can try it at speedprompt.app. It’s free, it’s early, and feedback at this stage shapes everything that comes next.

The prompt box doesn’t have to feel like a blank page every time. That’s the bet.