CV Match: Why I Built It, What I Refused to Build, and What I Learned After 100 Real Users
A Small Number, a Real Moment
CV Match has been out in the world for a few months now. I almost didn’t notice when it crossed 100 installations. That number, by itself, is not impressive. It is small. Almost trivial.
What made me stop was not the number, but the realization that this thing actually existed in the hands of people I did not know. Real users, not personas. Real decisions, not demos.
When AI Promised Clarity and Delivered Noise
I have been circling around versions of this idea for years. At least once, I even started building something similar and abandoned it halfway through. That time coincided with the moment Microsoft went all in on OpenAI. It was obvious that LinkedIn would follow. I expected that moment to finally produce tools that made job searching clearer and more honest. Something that reduced noise instead of adding to it.
What actually arrived was depressingly predictable.
More closed experiences. More premium gates. CVs you could not really control. Chat interfaces where structure was needed, and marketing language where precision mattered. The tools looked modern, but the experience felt as murky as ever.
The core question remained unanswered.
The Question That Would Not Go Away
At the same time, my wife was, and still is, looking for a job.
Watching that process up close reactivated a feeling I knew well from my own searches. Not anxiety, exactly. More like a constant, low grade uncertainty that never quite goes away.
- Am I actually a fit for this role?
- And if not, where exactly is the gap?
Not in a motivational sense. In a practical one.
Is this a small adjustment or a fundamental mismatch?
Which skills are missing, and which ones are close enough to stretch?
Is this worth the effort, or is it just noise?
CV Match started as a way to make that question answerable. First for her, then as an experiment to see whether others were struggling with the same lack of clarity.
They were.
Drawing Lines Before Writing Code
Before writing any code, I spent time deciding what I was not willing to build. This part mattered more than feature ideas.
I did not want another tool that quietly harvested CVs. I did not want user data stored “for improvement.” I did not want subscriptions tied to anxiety. I did not want a system that locked people into one provider or one opaque workflow.
So I chose constraints:
- Your CV never leaves your device
- You bring your own API key
- You choose the model and the provider
- The cost of using the tool is visible and trivial
- The output has to be structured, grounded, and useful. Not chatty. Not performative
Those constraints made the product smaller. They also made it honest.
What CV Match Actually Does
CV Match does not try to be clever.
You give it a CV and a job description. It tells you how well they match, and more importantly, why. It points out missing skills explicitly. It suggests concrete edits you can make. If you want, it generates a tailored version of your CV you can actually download.
That is it.
No scoring theater. No vague encouragement. No pretending the decision is binary.
What People Thought They Wanted
Before releasing it, I asked people what they thought they would want. The answers were revealing.
Most cared about actionable edits. A raw match percentage was not enough. Missing skills only mattered when paired with next steps. Downloadable CVs were useful, but not the core value.
That distinction stuck with me.
People are not looking for judgment. They already get enough of that. They are looking for orientation.
Reality After Shipping
After a few months, reality looks modest but real.
A bit over a hundred installs. No paid marketing. A handful of organic posts. Most users coming from the US, even though I am based in Israel. A surprisingly high share of ChromeOS users. And very few uninstalls.
This is not product market fit. Not even close.
But it is confirmation that the problem exists, and that the way I framed it does not repel people on contact. That alone is valuable.
What Building It Taught Me
The build itself challenged a few assumptions I had.
Structured reasoning mattered more than model branding. Generating useful edits was significantly harder than generating analysis. Small changes in prompt phrasing had outsized effects. And people were extremely sensitive to tone when it came to their CVs.
That last point mattered most. It reminded me that language is not decoration. It is the interface.
What I Am Not Rushing to Do
There are obvious directions I could take this next. I am deliberately not rushing into them.
- No growth hacks
- No urgency driven monetization
- No tricks designed to exploit insecurity
If this grows, it will be because people trust it, not because it pressures them.
Why This Project Mattered
CV Match was never about scale.
It was about finishing something. About respecting users. About grounding myself again in first principles. About building a small, focused tool that does one thing well.
Side projects are where you remember what real product work feels like, before layers of abstraction and politics creep in.
If you want to try it, CV Match is here.
And if you have feedback, especially the uncomfortable kind, I genuinely want to hear it.