It all started when I decided I wanted to own a giant decorated egg.
Last month there were over 100 eggs being auctioned off for charity, each painted by a different artist, after being displayed across London (and somehow, miraculously, not vandalised).
I wanted one. But which one to bid on?
My grandad always used to say: if you’re going to own an inconveniently large egg, make sure it has the potential to be worth something one day.
So with that in mind, I turned to AI-assisted coding.
First, I used Cursor to write a script that scraped the URL of every listing, extracted the artist name, and filtered down to those where the current highest bid (if any) was £500 or less.
With those logged to a spreadsheet, I then asked ChatGPT to develop a system for evaluating an artist’s commercial potential. It came up with an elaborate scoring system including where they studied, any prominent exhibitions they featured in, and public sales records of their previous work.
Back in Cursor, I had it write another Python script that would use the ChatGPT API to iterate through each artist, running web searches to assess each contributing factor and giving each a final “commercial potential” score.
Finally, given that I’d need to shout “STAY AWAY FROM THAT!” at our kids 36 times per day for the next 15 years, I wanted to make sure it was something we actually liked. So my wife and I went through the top scorers, and identified our favourite: Jay Kaes, with a ChatGPT-approved score of 85/100.
And this is where an already pointless story takes an even more disappointing turn.
Interest seemed generally low, so I placed our bid just before going to bed ahead of the next day’s 9am deadline – pretty confident that no-one would snipe in at the last minute.
But they did! Which came as something of a relief, because by then I’d properly thought through just how big these things are.
In one final twist, the winner pulled out and I was offered the egg a few days later as the next highest bidder. But by that point I’d lost interest in the whole thing.
I’m now thinking that £500 and a lifetime of smash-related anxiety would have been worth it to have a photo and a more satisfying end to this story, but hey.
The point is: I figured out how to do all this in less than an hour one evening. It’s truly incredible what non-coders can now do – and while it’s probably better applied to something less monumentally pointless, having little “I wonder if…” personal projects like this is a great way to learn.