How I Built a Search Tool for DJs Without Knowing How to Code
By Kushal Patel, Innovation Manager
A few months ago, I didn’t know how to code.
I still don’t. Not really.
But in that time, by finding some rare moments of peace in the insanity of living in London, I’ve managed to build a simple web app with the help of AI coding tools. And through my experience, I’ve come to deeply appreciate the power of these tools and get a glimpse into the impact they could have on the music industry.
My goal with this project was to solve a real (if very minor) challenge I face regularly as an amateur DJ. It’s a tool no company would ever bother making, because the problem is too niche. Too boring. Too small.
But it mattered to me. So I built it. And all without writing a single line of code.
In this article, I’ll explain the (not completely painless) journey I went on to build the app and some of the key lessons I learnt along the way. From the enduring power of good ‘ol fashioned research, to the impact this technology could have in realms as seemingly distant as artist marketing, I have finished this journey feeling more invigorated and knowledgeable about the potential of AI.
You’ll even get a chance to try the app for yourselves, and for the DJs amongst you who knows, even use it for yourselves in the future.
The Problem: Searching for Downloads is Painful
When I want to download a new track to mix with, I usually have a preferred order of places to check:
My first choice is Bandcamp. Why? Because it’s a platform where a large proportion of my purchase fee goes directly to the artist or label. And I like supporting artists, it makes me feel better about the world.
But sometimes, I’ll hear an edit of a track that blows my mind through its unique reimagination of a well-known sample. This Spice Girls inspired Trap banger should explain what I mean (credit Smochi, and of course the Spice Girls).
Now I know from many years working in music, that the likelihood of an edit like this being kosher in clearing the samples it uses is… low.
But nevertheless I still want to throw the track into a set. In this case, Soundcloud is usually my best bet because these bootlegs are often available as a ‘free download’ in exchange for following the artist on socials. And that’s because the artist is unable to monetise the track on streaming services due to the uncleared sample (there’s a million other nuances to the legalities here but I won’t get into those).
And in those cases where neither Bandcamp or Soundcloud has what I’m looking for, I’ll usually (reluctantly) go to Amazon.
My gripe? I end up having to search each site manually, opening multiple tabs, and pasting the same search terms over and over. It’s a minor hassle, but when you do it often enough, it adds up.
This was the kind of problem no one else would solve — but it was just annoying enough for me to spend some time on it.
The “What If” Moment
When I was first introduced to AI coding tools (specifically Cursor) by my boss, I immediately started thinking: Could I actually build something that fixed this problem… even if I didn’t know how to code?
The answer, at first, seemed to be yes.
Within about five hours of experimenting with conversational AI prompts, I had a working prototype. It was basic, but functional. I could type in a track name and get links to search results across Bandcamp, SoundCloud, and Amazon. All in one place.
I even managed to make it look pretty.
My opening prompt in Cursor
At this point I was ecstatic. The possibilities seemed endless and I was singing Cursor’s praises to basically anyone willing to listen (and a whole bunch of people who weren’t)…
This was my first major learning.
Lesson 1: Generative AI wasn’t just about chatbots (e.g. Chat GPT) and content generation (e.g. Dalle or Suno). It had a much deeper potential that I’d only just started to fully appreciate.
And given Cursor’s $900m investment round in June 2025, apparently I’m not the only one who sees this potential…
When the Magic Wears Off
Unfortunately this moment of euphoria was short lived.
For the next (I don’t want to think about how many) hours, I got stuck. Painfully and agonisingly stuck. I went down countless rabbit holes that led to know nowhere and ultimately made little to no progress on the app I’d spent just 5 hours getting to a working prototype.
In this period I was mainly focused on ‘deploying’ the app, which essentially means taking the code which was running locally on my laptop, and getting it live online for anyone to use (‘deployment’ is just one of many terms I had to learn the meaning of).
And I tried everything. I tried multiple tools (e.g. Vercel and Heroku), attempted to convert the entire codebase from Python to JavaScript (I was given some advice that it might make it easier to deploy), and at one stage even enlisted my Dad (who has almost 40 years of experience in software development). But none of it worked.
My rose-tinted spectacles were off and smashed to smithereens at this point, but perhaps unknowingly I had learnt another important lesson.
Lesson 2: The thing about AI-assisted coding is: when it works, it feels like cheating. When it breaks, and you don’t understand the underlying code, it’s brutal. You’re effectively debugging blind. It’s not enough to be able to prompt well, you also need to know how app development actually works.
Doing Some Good Ol’ Fashioned Research
So what did I do? I went back to basics. Back to the working prototype I had in Python. But this time, I was armed with a little more knowledge.
I’d spent some time learning about coding fundamentals the good ol’ fashioned way – articles, online courses, and asking people for advice. To give you an important yet incredibly boring example, I had to learn about Git and Github, which are used for versioning (which came with another set of terms I had to learn about). Crucially for me, these tools make it much easier to return to previous backups when things go wrong.
In addition, all that time I’d spent arguing with ChatGPT and Claude had taught me how to ask the right questions and troubleshoot more effectively with AI. Growing up in the age of Google, the learning tools at my disposal were on a scale unimaginable ten years before I was born. AI today seems to offer learning opportunities on a scale that was unimaginable even five years ago. The ability to bounce ideas back and forth, and troubleshoot conversationally is incredible. It took me a while to understand that this really is the way to use LLMs, but now that I have, my learning has accelerated much more quickly.
The turning point was when I put the following prompt into Cursor:
“How would you suggest deploying this app. I want it to live on a url that anyone can use. I want it to be as easy as possible to push updates from the local version of the app to the deployed version.”
Now I have no doubts that there are better prompts out there than this. But compared to my early prompts (which were broad and completely devoid of any technical language), this one had more targeted instructions. This ultimately resulted in more helpful answers.
The AI suggested I use a tool called ‘Render’ and I took its advice. Less than an hour later I had a working app! The sense of euphoria was back. I had just built a working web application without writing any code whatsoever, a prospect I didn’t even know existed a year earlier.
But I hadn’t forgotten all those hours I’d spent hitting walls and trundling along paths that led to nowhere.
All in all, this experience had ultimately taught me my third major lesson:
Lesson 3: The deep knowledge that development/engineering teams have is going to be as, if not more vital, at least in the short-term and likely in the longer term too. When it comes to AI coding, aspects such as bug fixing, robustness, scalability, performance and security are just some that I think will be even more important to consult expert developers on. Getting to something that works will be easier than ever before. Getting to something that works consistently at scale might be just as hard as it ever was.
Much to my disappointment the sphere of AI coding has been labelled ‘Vibe Coding’ by the internet. Other than being painfully ‘trying too hard to be cool’ I also think it implies that you can build apps with no knowledge of how technology is developed. At least at present, that is simply not true. It’s only by learning about how apps are built have I even been able to get this far.
So What Did I Actually Build?
Well rather than explain it, why don’t you have a play for yourself? Type a track into the search box, or put in a streaming link. See what happens.
Click here for demo: https://search-for-djs.onrender.com
Maybe it breaks, maybe it returns nothing, maybe you don’t see the point of it all. But maybe, just maybe, this tool could make your life just a little bit easier. And whichever one of these describes you I’d love to hear your feedback. I’m usually pretty responsive on LinkedIn, so shoot me a message on there: https://www.linkedin.com/in/kushal-v-patel/
And if you don’t want to try it yourself, here’s a couple of short videos that explain some of the main use cases.
An edit on Soundcloud where a free download is available:
The Bigger Picture: AI Coding and the Music Industry
I spend a lot of time researching AI’s impact on the music industry – in my day job and in my spare time. Over the past few years, there have been moments that have really made me stop and think. Moments which have made me look at the world with equal parts excitement and fear about how much AI will change the way that we live and work. No doubt some of these moments have come from the tools that everyone talks about (Video generation with Veo, Image generation with Dalle, Music generation with Suno etc.). But Cursor was up there with the most mind blowing.
In my humblest of opinions, the conversation around AI’s impact on the music industry has so far focused too heavily on Gen AI Music Creation tools like Suno, Udio and Riffusion. And whilst that is a fascinating area with some very important existential and moral questions about how music is created and protected, I think the real benefits of AI to the music industry will come from the much less sexy stuff. Automated workflows, data analytics, payment processing, and yes you’ve guessed no-code development.
Imagine a world where a label’s marketing team can spin up interactive websites, promo tools, remix contests, and fan activations at a fraction of the cost.
Or where an artist could use AI to analyse which of their tracks were performing best with different segments of their audience, and then automatically use that to optimise release dates, setlists, or content strategy.
Or where a DJ can build the exact tool they need to streamline their curation flow.
It’s in these areas that AI will quietly change the game — not by making music, but by making artists and their teams more powerful than ever before.
Note: The views expressed in this article are my own and do not reflect those of MassiveMusic