Prompt Builder

Body:

Hey IH 👋

I'm a developer who's been deep in prompt engineering for the past year, and I want to share a problem I ran into and the tool I built to solve it.

The problem

When I first started writing prompts, they were simple. A sentence or two, maybe a paragraph. But as I started doing more serious work — building AI-powered features, automating workflows, testing different models — my prompts grew. Fast.

A single prompt would have a system role, task instructions, constraints, tone guidelines, output formatting rules, and few-shot examples. Some of mine were pushing 1,000+ words. And I was managing all of this in a plain text box or a Google Doc.

Every time I wanted to test a small change — like adjusting one constraint or swapping the persona — I'd duplicate the entire prompt and carefully edit one section. I'd lose track of which version was which. I'd forget what I changed between versions. It felt like writing an entire application in a single file with no functions and no structure.

I kept thinking: developers solved this problem decades ago with modularity. Why are we still treating prompts like monolithic blobs of text?

What I built

I built Prompt Builder — a block-based editor where you construct prompts from independent, reusable components instead of writing everything as a wall of text.

The core idea is simple. Each section of your prompt (role, instructions, constraints, examples, output format) becomes its own block. You can drag blocks to reorder them, toggle them on or off to test their impact, tag them with custom XML structure, and reuse your best blocks across completely different prompts.

Some of the features that have been most useful to me personally:

A variables system that lets you create dynamic templates. Write a prompt once with placeholders, then swap in different values without rewriting anything. This alone saved me a ridiculous amount of time.

Show/hide toggles on individual blocks. This turned out to be the killer feature I didn't expect. Want to know if your few-shot examples are actually improving output quality? Hide that block, run the prompt, compare. It makes A/B testing prompt sections dead simple.

Multi-model execution so you can send the same assembled prompt to different models and compare outputs side by side.

And editable AI responses, which lets you refine what the model returns and feed corrected outputs back into your iteration loop.

Where I am now

The tool is live and functional. I've been using it daily for my own work and it's genuinely changed how I approach prompt engineering. But I'm at that uncomfortable stage every indie hacker knows — I built something that solves my problem, and now I need to find out if it solves other people's problems too.

I haven't done a big launch yet. I'm not running ads. I'm just trying to get the tool in front of people who do serious prompt work and hear what they actually think.

What I'd love from this community

I'm not here to pitch. I genuinely want to know:

1. Does this problem resonate with you? Have your prompts gotten complex enough that managing them feels painful?

2. If you tried the tool, what's your honest first impression? What's confusing, what's missing, what would make you actually use this regularly?

3. How are you currently managing complex prompts? I'm curious what workflows people have cobbled together — maybe there's something I should learn from or integrate with.

I'm also happy to share anything about the build process. The stack is Next.js 15, React 19, Supabase, and Zustand. I've learned a lot (and made a lot of mistakes) along the way.

Thanks for reading. I know this community values real talk over hype, so hit me with whatever you're thinking — I can take it. 🙂