Going from years of casual coding to shipping a full-stack financial app, and why I avoided AI.
My degree is in Zoology & Mammalogy, not Computer Science. But i’ve been working with code on and off for years. Starting with making silly indie games when I was younger, and later messing around with algorithmic trading systems. Even with that background, jumping from writing code for fun to shipping a live proprietary backtesting app (MacroStraty) with custom Python scrapers covering 10+ years of economic data, a full front and backend, auth system, and payments was way more work than I imagined at the start.
I actually built the whole thing originally as a personal project just to help with my own trading. At the start, the obvious temptation was to let AI generate the entire codebase because why wouldn't you? It's faster and easier.
Instead, I kept AI out of the core code-writing and developed the architecture myself using the knowledge I'd picked up over the years. I had the free time, and honestly, I just missed the feeling of building something big with my own hands. There's a unique kind of pride you get when you figure it out yourself I find (the exact same feeling I used to get coding games as a kid).
Where AI did shine for me however, was acting as a debugging tool and a place to bounce ideas off when figuring out how to link different parts of the stack together. Letting it help handle the brainstorming and troubleshooting while I actually wrote the logic gave me the best of both worlds. I know exactly how all the code works and how every element is intertwined (making it easy to find where problems are).
It was definitely slower early on, and there were plenty of moments I was tempted to just let AI take over completely (and to be completely transparent, there are 100% some parts where I had absolutely no clue what I was doing and did enlist the help of AI to get things moving again), but sticking to my guns really paid off when things got harder. The real bottleneck I found wasn't building the UI or linking pages or anything; it was the tedious, data-specific problems:
Handling timezone shifts across 10+ years of global economic releases (GDP, CPI, rate decisions).
Hard-locking historical data points so future data never leaks into backtests.
Writing resilient scrapers that don't constantly break, plus a solid alert system for when they inevitably do.
Getting the built-in trading simulator to actually read the OHLCV candle data and run like a real chart (I was stuck on this issue for MONTHS).
Because I wrote the backend myself and only really used AI to help debug and connect the pieces, fixing those edge cases personally took a significantly smaller amount of time than if I had no idea how any of my software worked and was going in blind. I mean, specifically on the trading simulator issue, I did at one point try and get AI to develop it for me but it kept going round in circles and just couldn't figure it out so I had to do it myself anyway.
For other self-taught or non-CS builders trying to ship technical products right now:
How do you split the work with AI? Do you let it write the bulk of your code, or keep it more as a debugging and brainstorming tool like I do?
If relevant to you, what was the biggest change for you going from small-time coding to running a fully fledged software with paying users who rely on you?
Also, is there any advice you can give to someone starting out on the marketing side of software? Is AI actually helpful for making engaging ads or growing social media accounts or should it be avoided?
Replies
Be the first to reply
Have a question or a thought to share? Add a comment above to start the conversation.