How bad does a team have to be to inspire you to start building a web app?
For me, at least, the answer appears to be Group-Stage Czechia at the 2026 World Cup. As someone from the US with Italian and Czechoslovakian (yes, both countries, hence I use the old name still) heritage, there haven’t been many teams I have some kind of familial tie to at the past several (men’s) World Cups, so it was nice to be able to root for Czechia alongside the US and the several other nations I tend to go for in international sporting competitions.
By minute 20 of Czechia’s first game - down in Guadalajara, Mexico - it looked to me like the central Europeans couldn’t really hang in the humid climate quite as well as their South Korean opponents could. This got me wondering: is the geographic scale of this World Cup going to give some nations an advantage over others?
The 2026 World Cup is far and away the most vast ever orchestrated, covering basically the entire North American continent. It’s sometimes lost on the citizens of other nations just how vast North America’s Big 3 constituent countries are - in particular Canada and the US. What could be a beautiful, breezy day on one side of the country/continent can be a humid, swampy, and oppressively hot evening on the other side. Add to that a tournament schedule that can see a team traverse the breadth and/or length of the entire North American continent - maybe more than once - and you’ve got a recipe for a physically challenging endeavor before anyone even sets foot on a pitch.
This got me curious: could I map out where each team was going to be playing, what the weather might be at their destination, and how this might impact their ability to win any given match? It seemed like a great project to really dive into some of the AI coding work I’ve been getting into in my spare time, so I jumped into Claude Code and got into it. The original concept came together quite quickly, largely drawing from open-source data repositories such as football-data.org, OpenStreetMap, and Open-Meteo. Was it anything I couldn’t do “by hand” on my own? No, but my general attitude towards agentic coding has been that, 99% of the time, when I’m doing the coding myself I end up turning to Google/StackOverflow every 5 minutes or 5 lines (whichever comes first) anyway, and this is, essentially (that word’s doing a lot of work, I know), what Claude Code is doing too, but it can type a lot faster than me. For what it’s worth, there’s a whole bunch of more more niche code I work with, largely tied to my PhD thesis, that AI has failed very miserably at, so I definitely don’t recommend using coding agents for anything that’s not very “common” in nature with respect to software development.
Now, as anyone who’s done some agentic/vibe/AI/etc. coding will tell you, it’s definitely addictive (especially addictive, I’d say, for anyone who’s actually done a decent amount of software development work previously), and I’m no exception. I started wondering what else I could incorporate into this “dashboard” to make it a one-stop shop for all things World Cup (at least for me, personally). First I added more metrological data and overlays - temperature, radar, pressure, alerts - and then incorporated live (realistically semi-live, I’m not paying for premium API access) score data.
From there, I started wondering how I could incorporate other AI-based tools I’d come across. Two, in particular, that I thought could be interesting to play around with using my World Cup dashboard as a test case were MiroFish and OpenAI’s Symphony, incorporating a predictive/modeling element to the dashboard (and drawing from dashboard data sources) as well as bringing feature management out of the Claude Code CLI and into a Kanban-style interface that would let me manage and track specific dashboard features, respectively. Interested in pushing the envelope of what could be done autonomously, my initial incorporation of both repositories’ codebases consisted of pointing Claude Code at the repo URLs and giving a rough description of what I was looking to draw upon for each to incorporate back into the dashboard site. I won’t say it was perfect from the get-go, but once Symphony, in particular got up and running, it became very easy to incrementally improve and adjust the World Cup dashboard and all of its constituent features - shifting from a “developer-style” interaction with a CLI-based coding agent, to a “project manager-style” interaction with a Kanban board that automatically called up coding agents for new features/adjustments as I thought of them and entered new “tickets”. (Brief side note, all of this is running on my personal web server hosted through DigitalOcean.)
Suffice it to say, it’s pretty surreal to sit in a role akin to a project manager for a software dev team where all the “team members” are just instances of Claude Code running on your personal web server. Also suffice it to say that, if I was trying to do something that went much more “against the grain” with respect to software and web app development, I don’t think I’d be having so much success with the “coding” interface I’ve settled into. Nonetheless, the app is live, it works decently well, in my opinion (I’ve certainly been checking it multiple times a day to stay abreast of what’s coming up in the tournament), and it’s been borderline trivial to incorporate ever-more complex features and behaviors (my favorite part to play around with is the sections of the web app that yet you project and map out different scenarios and possible outcomes - something I’ve done a bunch of recently following several rough losses for teams I was rooting for). Do I think AI is coming for all our jobs? No. Do I think AI is going to replace software developers? Probably not. Do I think there is a set of work that software developers used to do that is now going to be taken up by AI instead? Yes. This “boilerplate” kind of work certainly feels like it’s not going to be done by a human anymore. Do I think it’s easy to maintain a strong and consistent “brand identity” or establish truly compelling data visualizations with AI coding agents? Not at all.
I wouldn’t say my World Cup dashboard is particularly well branded, or even designed, but that wasn’t what I was going for initially, at least. First and foremost, it’s something I made for me, to help me answer questions I was having about the things I was observing, and to provide a tool for exploring what could have been. That being said, it is live, I do encourage you to play around with it and, if you are so inclined, to check out its GitHub repo as well. I do think it’s come together well enough over the course of this tournament to warrant a more thorough dive on my part (the women’s World Cup is coming up and then the 100th anniversary World Cup in Iberia/Morocco), and I’m excited to continue using it as a platform to play around with all the interesting tools coming out for both humans and AIs to develop and deploy software.
Before I go, I’ll leave you with the link to the github.io page for the dashboard, from which you can also get to the rest of the web app hosted on my server: https://joeborrello.github.io/world-cup-2026-dashboard/
Have fun browsing, and enjoy the last few days of this 2026 World Cup!