The other day I decided to see how far I could launch the worst programming interview question in existence: fibonacci. You give me Claude, I'll give you more fibonacci than you can handle.
There's no script for this one and no plan. Just me, a dumb interview question, and having my mind blown yet again. What came out the other side wasn't a monolithic script, it was a modular library that treats the sequence as a lens for looking at architecture, at the underlying maths, and at how something looks when you actually care how it looks.
I fed the result to Gemini for a read, and it came back with this:
This submission is an exemplary piece of exploratory engineering. It moves far beyond the "standard" Fibonacci implementation, treating the sequence as a lens through which to demonstrate architectural discipline, mathematical curiosity, and high-end visual craft. Instead of a monolithic script, the candidate delivered a modular library that balances theoretical rigor with pragmatic software design. This is more than a solution to a problem; it is a demonstration of how a senior engineer approaches discovery. Recommendation: Hire.
Which is a very fancy way of saying that an interview question everybody rolls their eyes at turns into something genuinely interesting the moment you stop trying to answer it and start trying to explore it.
All the code is up at github.com/robconery/fib-with-claude if you want to poke at it. I learned something building this, which is the whole reason I keep doing it. Enjoy the video.