I started with a spreadsheet because that was what I knew. It could pull market data, and I could work with formulas. For someone who wasn't a developer, it was the obvious place to begin.
But as I tried to do more with it, my laptop choked. The idea still needed testing; the tool I knew was getting in the way. Somewhere along the way, I discovered that people typically use Python for this kind of work.
That explained what I needed, but it left me with a different problem. I didn't know Python either. Moving from a spreadsheet to code would only help if I could actually build the thing.
Until then, my use of AI had been fairly ordinary. Summarize this document. Search this for me. Create an itinerary, write a mail, format something. Basically glorified search. Useful, but it hadn't changed the kind of thing I could attempt.
Now I had a problem where those uses wouldn't get me very far. I needed code that could fetch historical data and test my ideas. AI was still new, with the promise of doing wonders, so I decided to try my luck.
To my surprise, it helped me write the code. That got me past the barrier of not knowing the language. But this was well before today's agentic workflows. Writing the code did not mean the whole thing was done. There were bugs to fix, and a lot of back and forth before the pieces worked.
Through that back and forth, I gradually built a collection of Python scripts. They could fetch the data and let me test ideas in seconds instead of spending hours doing things manually. The original question was finally something I could investigate.
Those early scripts eventually became a dashboard I still rely on today. The investment strategy luckily turned out to be a worthwhile exercise. But what stayed with me was the experience of getting there. This was probably the first time AI had helped me build something I simply couldn't have built on my own.
That mattered beyond this one project. Ever since graduation, I had ideas I never even started because they felt too big. They needed capabilities I didn't have, especially coding. Having an idea was easy enough. Getting from that idea to something I could actually use was where I kept stopping.
The investing project gave me a reason to try those ideas again. I had already gone from a tool that couldn't handle the work to a language I didn't know, and still ended up with something useful. I started working on other applications for myself.
Some of those applications are scrappy versions of software I would otherwise have to pay for. They only need to work for one person: me. That makes the scope much smaller. I can build for my use case without taking on everything the full product does.
So my use of AI grew out of a problem I already cared about. The scripts helped me explore investing ideas; building them made other personal projects feel possible. Alongside that, I use a personal agent for repetitive things such as ordering groceries and renewing insurance. Some uses save time, some save money, and some let me go after an ambition I had kept putting off.
People often ask me whether paying for AI subscriptions and tokens is worth it. The momentum project is still the example that comes to mind. That one project has already paid for the tokens that went into building it, and probably quite a few future tokens as well.
I started with searches and summaries, but the change came when I had something I wanted to do and couldn't get it done with the skills I had. The spreadsheet reached its limit. I couldn't write Python. AI helped me keep going. I still don't know Python, but I now have a dashboard I use and a much longer list of ideas I am willing to attempt.