AdLuv was the first product I created as a vibe coder.
It's a tool for discovering, tracking, and organizing ads across LinkedIn, Meta, and Google. But this article is mostly about what I learned while building it.
A lot of professionals have a love-hate relationship with AI, especially when it comes to writing code, but I decided to slowly embraced it. I wanted to see what I could create with it and how much of the work I could actually get done.
I'm the man of 1,000 ideas and 0 free time. AI helped me turn the vision in my head into something clickable in a really short time, without a full team around me. That was enough to make me want to keep going.
Jack Oldham was my co-founder at AdLuv. He created the initial plan, and I fed it to the clanker. I had an idea of what the product should do, but I hadn't thought through the implementation details well enough. Feeding the plan to the LLM wasn't enough on its own.
The LLM created something that looked good, then I showed it to Jack Oldham, the marketing brain behind the product, and the UX didn't make much sense. The screens were there, but the way someone would actually use the product needed more work.
So I went through the part I enjoyed the least: collecting Jack's feedback, sending it back to the LLM, reviewing the changes, and doing it again.
That took longer than I wanted and I hated it. Getting the first version on the screen had been fast, but getting it to make sense was slower.
I learned to vibe code in smaller iterations. Build a part, review it, test it, then move on. It's easier to control what changes, understand what's working, and deploy something you have actually checked. Giving a weak plan to the clanker is the worst mistake you can do.
In the end, I shelved AdLuv because it didn't bring anything new to the table compared to the competitors. Selling it cheaper to grab a little market share was an option, but I didn't see that turning into consistent revenue for me. I needed a better reason for someone to choose it.
And dealing with the Ads directories trying to block you, rotating proxies, solving captchas, was something I was already doing on a daily basis; I didn't want to have that after hours.
I was mostly happy with the result, even though I hated the review and feedback part. I got to turn an idea into a working product and learned a few things about how I want to build the next one.
I've decided to put the code out there. Below are a few screenshots of the product.
Ad library
Advertiser overview
Ad details
You can find the source code on GitHub.