With Claude Code getting more and more popular, I wanted to see what and how quickly can be done development wise so I decided to make running database website excusively developed and generated by AI from the scratch and that’s how Runelio was built. So let me run you through a quick story.
Running in Barcelona
As I am living partially in Barcelona for the last few years, I started to run a lot. It is hard not too as “half of the city” is running. Wherever you go, runners are around you. Beach, avenida Diagonal, small streets… You name it. Great weather, beach city, wonderful arhitecture, nation with great sports culture and infrastructure. Obviously, not had to get motivated to start running even if running is not your favorite activity.
I started running some races, mostly 10K and even some half marathons. As there is race in Barcelona or somewhere around in Catalunya almost every week, you have a lot options to participate and feel the energy. Most of the races are perfectly organized and it is amazing feeling to participate in them. To this day, races in Barcelona and energy around it, are still one of the my favorites in the city and I advise everybody to try it.
As I got into some races I noticed that even 10K races usually have lot of participants, some of them more them ten thousands. I also noticed that lot of people are quite fast and fit so being analytical guy that really likes numbers, I wanted to analyze differerence in finishing time between races in Barcelona/Spain and Croatia and compare number of finishers.
Idea was born
While you can do it manually, I wanted to see what AI can do. I wanted to get introduced to vibe-coding and see how far this can go. So the initial idea was born. Let’s try what AI can do and make a website that will be able to compare some races with imported data. Import few half marathons and 10K races and compare it with Zagreb half marathon and 10K. Important to note that I started with this project in March 2026, in the meantime AI models improved.
While I definitely do have some tech background and understanding, I am far from being developer. The last time I did some real programming was 9-10 years ago and that’s when I was using PHP and was trying to parse some data for basketball analytics. It took a lot of time back then so I was curious to see how long will it take to parse data, prepare website and do some analytics around it.
Four hours later
For the programming part, I was using Claude. And later, for any product ideas etc., I was using ChatGPT. That was by design, I wanted to use both of them. So I started writing prompts and user stories in Claude about what do I want to do. I used Sonnet 4.6 model. I was strictly following instructions that Claude provided. I had to install lot of additional software to my Mac, but it was always very easy as Claude is always providing info on what to run in terminal. And if there was an error, I just pasted it to Claude and within one or two prompts, solution or new file was generated.
Long story short, in less then four hours I did the following, from the scratch:
- I parsed two races, Barcelona half marahton and Zagreb half marathon
- I had website running locally
- I added lot of different stats so the races can be compared
- I had UI with six different languages
The results are the following:
So yeah, as you can see, runners in Barcelona generally have much faster results then the one in Zagreb, confirmed. Not to mention numbers of finishers (40k participants registered for 2027 half marathon in five days).
Once I got these two races compared, I become greedy and I added added few more half marathon, marathon and 10K races to compare. I was mostly adding popular (half) marathons in Spain and 10K races from Catalunya and comparing it.
For any race along with being able to see results and search for runners, I was getting summary:
Later I added weather and humidity info at the start of the race:
Within like eight hours of work, I had a local website built with next.JS running on my Mac, around 20 races imported to Supabase free instance, lot of stats generated. The majority of time went to parsers generation because practically lot of races had a different results format so it required plenty prompts to Claude, inputs by me and some QA-ing to check that everything is imported properly (for example gun time and chip time). If was a bit challenging sometime to find input Claude needed for parsers and I had Google Chrome network console opened. Not all the sites have easily accessible API’s and some of them where showing only gun time even though by clicking on the runners name in the results page you could see the chip time, the one that I needed. But after few prompts, I got what I needed. In any event, I was impressed by what I can do.
For comparison, 9-10 years ago, when I was doing scrapers for Euroleague games, it took me several days to make one parser. Not to mention that it took me a lot of time to parse all season games as I had to design database, run the scripts manually for every game etc. And then make a website to view all the data, calculate standings, stats etc.
MVP
So after proof of concept, I didn’t do anything for the next few months on the running website until this summer arrived. Being back to Zagreb over summer, I got motivated. There is no website that has results of all the races in the world. So I wanted to make an MVP and gather as much results from Croatia and Spain and make a database and then publish it online.
I used ChatGPT to get some feedback on what to do. I can say ChatGPT is amazing for discovering ideas and proposing MVP solutions and additional features. And with new models (last one I used was Sol 5.6) it just got better. It really gives plenty ideas if you prompt correctly.
Along with that, with the help of OpenAI tool, I decided that the website will be called Runelio. I assigned ChatGPT task to give find me proper name, the one that is kinda fun, with free domain. Unfortunatelly, I didn’t think about Instagram at the time, so I had to use a bit different nickname for it. Also I chose one of many slogans ChatGPT proposed, the one that isis kinda fun and marketable: “Run Your Story”.
I bought domain, published everything online (with the help of Claude of course), added around 200+ races and also created Instagram profile. Currently few posts and stories are already published and all the photos and texts are generated by ChatGPT. I just provide some stats from the website and after few prompts and attemps, I get generated photos that I want.
Also added leaderboard with race filters per country and distances so best results can be seen along with number of finishers.
Races with multiple editions got improved summary so users can track trends. Every race also has info on humidity and weather at the start of the race and in the summary you get average, median and highest temperature/humidity. For instance, Valencia maraton, one of the best in the world:
I am hoping to launch the website this month officially. The goal is to get as much races in the database as possible and to improve runner profiles so it can be useful. Also, running the agents that would automatically parse race results and generate stats as soon as the races are done so not a lot of manual work is needed by me.
Who knows what next is gonna come up to my or ChatGPT’s mind, but in any event it is very interesting project that would not be possible without Claude. Few years ago it would take plenty of people to build it, and now I can do a lot of stuff working few hours per week by myself.
Comparing to decade ago
On a side note, I mentioned that my last programming project was parsing data for basketball and that it took me few days to just make a parser and gather season data for one league. Few days ago I tried to do the same with Claude, but with Sonnet 5 model. I still didn’t try to code with Opus 5.5 not to burn plenty of tokens. I would of course use it if needed, but Sonnet was enough for me at the moment.
Within three hours I had:
- local website named HoopIndex with results of all Euroleague, Eurocup and ACB games for the last five years and associated boxscores
- player profiles with some advances stats
- player leaderboards
- team leaderboards
- job that parses game stats automatically after the game is finished
- six languages supported
Talk about increase in productivity…
AI revolution is in full swing
The conclusion is that we live in big revolution and the potential is scary. As a product owner or manager, you can now easily write a story and get a prototype for any idea. So you have lot of options to play around and you do not need to bug developers or designers to come up with some proof of concept. You can do it by yourself. You can also develop something, but the problem happens when you need to debug. I also noticed it myself during the project as the deeper I went into it, I had to run more and more prompts to debug and to finally get to solution. But it will for sure improve as the speed of AI model development is crazy.
And while everybody is scared of AI, I believe it will in the long run make our lives better, more productive and it will create jobs that we are not even thinking of today. So play along along and use it as much as you can.
