I built Promptimizer — a tool that turns bad AI prompts into great ones (and here's how I did it)
Hey everyone 👋
I'm Jaber Said, and I just shipped Promptimizer — a web app that takes your rough, messy AI prompts and instantly optimizes them to get better results from ChatGPT, Claude, Gemini, or any LLM.
The problem I kept running into:
I was using AI every single day for work, side projects, writing, coding — you name it. But I noticed something frustrating. Sometimes I'd get incredible responses, and other times the same model would give me garbage. The difference was never the AI. It was always my prompt.
I started studying prompt engineering, reading papers, testing frameworks like chain-of-thought, role-based prompting, and structured instructions. It worked — my results improved dramatically. But it was slow. Every time I wanted to ask something, I had to mentally restructure my prompt before hitting send. That felt like a problem worth solving.
So I built Promptimizer.
You paste in your raw, unstructured prompt. Promptimizer analyzes it and rewrites it using proven prompt engineering techniques — adding clarity, structure, context, constraints, and output formatting. You get a polished prompt in seconds that consistently pulls better answers from any model.
What it does:
It takes something like "write me a marketing plan" and turns it into a detailed, structured prompt that specifies the audience, tone, format, constraints, and expected output — all the things that make LLMs perform at their best.
The tech behind it:
I built this using [your stack — e.g., React, Next.js, Tailwind, OpenAI API, etc.]. The whole thing started as a weekend project, but once I saw how much better my own AI outputs got, I knew other people needed this too.
Where I am now:
I just launched publicly and I'm looking for early feedback. It's free to use right now and I want to hear what you think — what works, what doesn't, what features you'd want next.
Some numbers so far:
🛠️ Built in [X weeks/months]
🚀 [X] users since soft launch (if applicable)
💬 [X] prompts optimized (if applicable)
What's next:
I have a roadmap that includes prompt templates for specific use cases, team collaboration features, browser extensions, and API access. But honestly, what I build next depends on what you all tell me matters most.
I'd love your feedback:
Does this solve a real pain point for you?
What would make you use this daily?
What's missing?
Try it here 👉 promptimizer.top
Thanks for reading. Happy to answer any questions about the build, the tech, the idea — anything. This community has been a huge inspiration and I'm grateful to finally have something to share.
Let's go 🚀
Kai's comment about domain-specific templates is spot on. General "make any prompt better" gets harder to differentiate as models improve. But there's a middle layer that still matters: prompt structure itself.
Most people write prompts as one block of text. The model then has to figure out which sentence is the goal, which is a constraint, and which is context. That ambiguity is where outputs go wrong. Even a great model will misweight things when everything is jumbled together.
What I found works better than rewriting is decomposing. Break the prompt into typed pieces: role, objective, constraints, output format, examples. Each piece has a clear function. The model processes them separately instead of guessing boundaries in a wall of text.
I built an open source tool for this called flompt (https://github.com/Nyrok/flompt). Visual canvas where you drag 12 typed blocks and it compiles to structured XML. Different angle than Promptimizer since it focuses on the building step rather than the rewriting step. But the core problem is the same: unstructured input = unpredictable output.
Love this idea. You’re right — most of the time when AI gives a bad answer, it’s on us, not the model. But constantly “rewriting before you rewrite” in your head gets old fast. Having something that cleans up rough prompts in seconds is actually pretty appealing.I’m curious though — who do you see this being for long term? Power users who already care about prompt structure, or more casual users who just want better results without thinking about it? Feels like those are two very different audiences.Either way, congrats on shipping. Always cool to see someone scratch their own itch and put it out there.
Interesting idea. The pain point is real — I've noticed the same thing where the exact same question phrased two different ways gets wildly different quality from Claude or GPT. Prompt structure matters way more than most people realize. My honest concern though: how do you differentiate this long-term? The models themselves are getting better at interpreting vague prompts with each update. GPT-4 handles "write me a marketing plan" way better than GPT-3.5 did. So the gap between a bad prompt and an optimized one keeps shrinking. I think the real opportunity might be in domain-specific templates rather than general optimization. Like, a prompt template library for specific workflows — content writers, devs debugging code, recruiters writing job posts. That's where people would pay, because they don't just want a better prompt, they want a better result for their specific job. The general "make any prompt better" angle is hard to monetize when ChatGPT itself keeps improving at understanding intent.
Congrats on the launch, looks solid. How are you currently thinking about acquiring early users and gathering feedback?