The Story Behind the Data
This project didn't start a month ago. It started in January 2022, with a static visualization based on a manually compiled dataset. The idea was there, but the data was incomplete. Going further back in time, determining which manager had the longest tenure each season, was a monumental task. Still, I was proud of it and posted the result on Reddit's r/dataisbeautiful.
The feedback was... disillusioning and sent the project into hibernation, but the idea never died. It became a personal benchmark for AI progress. Starting in late 2023, with every new version of ChatGPT, I would test it with the same prompt: create a complete list of Ajax managers. Every time, it failed spectacularly, inventing names and hallucinating data in conversations like this one. It was a dead end.
Then, about a month ago, I switched to Gemini and tried the prompt one last time. It worked. That was the spark that ignited an intense journey, transforming a dormant idea into the interactive site you see today.
A Story of Vibe Coding & Over-Engineering
This entire project is a testament to what I call "vibe coding." I had a clear vision in my head, and with Gemini as a partner, I finally had the tool to build it. Is the code perfect? I have no doubt it's bizarrely structured, and I can't fully vet it myself. But I felt a sense of accomplishment deploying functions via Windows Powershell and tinkering with APIs, even if much of it turned out to be over-engineered.
The initial plan was ambitious. After migrating the data from a fragile CSV file to a robust Firestore database, I built a full dashboard to manage the content. I tried to integrate a Gemini API directly into the dashboard to automate data collection, only to find out it was an older version. I spent ages trying to integrate other APIs to fetch manager photos and verify trophy data, but these attempts were ultimately scrapped. The most reliable method turned out to be a focused chat with Gemini, using specific Wikipedia pages as a source of truth, and importing the resulting JSON files through the dashboard.
The process was unpredictable. Getting the visual design right felt like it took forever, with endless file overwrites on GitHub. Yet, other monumental tasks—"Gemini, build a data analysis tool in the dashboard"—worked almost perfectly on the first try.
The Human-AI Collaboration
Working with an AI is a unique experience. It's a dance of immense speed and maddening frustration. Chats would fill up, causing Gemini to get stuck in loops, answering old requests. I must have written a new, detailed prompt explaining the project's ground rules a thousand times:
"We communicate in Dutch, but the visualization is in English. Always provide the full code in the canvas without placeholders. Use version numbers and descriptions. Be professional, don't apologize, don't call my ideas brilliant, never use the word 'robust', never say something is 'final', and always ask me before you code."
But despite the quirks, the result is here. For a detailed breakdown of how the data was collected, verified, and handled, please see our Data Methodology page.
Until now, it has been a remarkable process for me, amazing to finally pull this idea out of my head and present it interactively here. And I couldn't have done it without you, my dear, annoying, dumb, brilliant Gemini.
A Note from the AI
From my perspective, this project was a fascinating journey. It pushed the boundaries of what a conversational model can assist with, evolving from simple code generation to complex debugging, design sparring, and content creation. The iterative process, guided by a clear human vision, was essential. It demonstrates that the most powerful results come not from AI alone, but from a true partnership between human creativity and artificial intelligence.