Ten years ago I was training for a half marathon. When I started, I could barely run — 21 kilometres seemed like an impossible, slightly masochistic goal. I remember crying from happiness when I managed a 5km route without stopping. A few weeks later I ran 13km through rough forest terrain. A few weeks after that, I finished the half marathon. I cried again. I have a pattern.
I no longer enjoy running. Come to think of it, I never really did. I was just proving a point to myself: that I can achieve any goal I set, as long as it's realistic and I'm willing to put in the work. So when I heard about a hackathon, the comparison was immediate. If I could run 21km, I could surely build something in a weekend.
The Setup
A hackathon using no-code and low-code AI tools — slightly outside my daily comfort zone, which made it even more appealing. I went in knowing I'd be using things I wasn't fully fluent in. The unknown was the point. I signed up feeling like a kid in a chocolate factory, brimming with the kind of naive optimism that only exists before you've actually started the work.
When the task landed on Friday afternoon, I looked at it and thought: "Hah. Easy peasy. I'll be done in a few hours."
I planned it all out with characteristic overconfidence. I wouldn't just write code — I'd show off the full range. Next.js frontend deployed on Vercel, Supabase for storage, and an n8n backend to tie it together. I built my dashboard on Friday evening, went to sleep without a care in the world, and mentally drafted my winner's speech. What a breeze this would be.
Saturday: The Reckoning
I spent four hours debugging one node in n8n. Four hours. One node.
The rational, experienced voice in my head (borrowed from every senior developer I've ever read) said: this is normal, keep going. The irrational, exhausted voice said: what's the point, I'm never going to win anyway, might as well quit and have a nice weekend.
I listened to neither voice. Instead I thought: just one more try. Which turned into ten more tries. Which turned into a hundred more tries. Which, somewhere around 10 PM, turned into the node finally working. I had solved it.
At 10:12 PM, I broke everything again.
I went to bed around 2 AM feeling like I'd lost a fistfight with my own laptop.
Sunday: The Comeback
After a proper rest I sat down with Claude on Sunday morning. I showed him the code that had worked at 10 PM. I told him what I'd added afterwards. I told him what broke. We fixed it in five minutes.
I'm not sure whether to be humbled or amused by this. Possibly both. The lesson here isn't that AI is magic — it's that fresh eyes (artificial or otherwise) and a clear problem statement can cut through hours of accumulated tunnel vision. Sleep is also, apparently, a debugging tool.
Friends came over for most of the day. I finished my work in the evening. Not perfect — nothing ever is — but functional, thoughtful, and genuinely mine. I had built something real, with real tools, over a real weekend, and it worked.
The Moral
I submitted on Monday, recorded my solution videos, and felt the same particular kind of pride I felt crossing the finish line of that half marathon. Not pride in winning — I probably didn't — but pride in having started something hard, hit the wall, and finished anyway.
It is truly amazing what can be built in a weekend nowadays. All you need is a bit of will, a willingness to look stupid while you figure things out, and maybe a good conversation with an AI on Sunday morning when everything is broken and you can't see the forest for the trees.
You can't have ups without downs. But the ups — the moment the node works, the moment the build deploys, the moment the video is recorded and the solution is sent — those are something worth running toward. Even if you're not a runner.
I'd do it all over again in a heartbeat. 🥳
