Back in university money was tight, I would check my finances every week to see how much I had to spend till the loan next came in. This fastidiousness continued after I started to work for understanding roughly where my savings would be, then later when I had quit my job to travel and needed to know how much more traveling I could do.
I always wanted a tool to automate this and make the output pretty. And back in university I wanted to learn proper programming, not just little scripts or html. So one summer I spent the time to learn Java, annoyed the shit out of Tom and made a cli tool which could only be described are barely functional dog shit. But it was good fun, I never used it after I created it, it just didn’t work.
From that point, I always wanted to go back and finish what I started with a full frontend. That was the dream. Luckily with the power of AI, getting these dreams complete is just so much easier now.
So here is FutureMoney. A simple free financial forecasting webapp where you set your accounts, schedule transactions and view the forecast. Some features this includes is:
- Local storage to not send any date off to a remote server, or a login to view your forecast from different devices.
- Scenarios to easily play out different forecasts.
- Native support for 100+ currencies. Your account can have a different currency to a transaction on the account, all of which could be forecasted in a 3rd currency.
- Graphical and tabular views for understanding the forecast.
- Transactions support one-off and repeating, with full customisation.
- The entire UI is easily re-arrangeable.
- Good old light and dark mode.
All data is inputted manually and updated manually. It doesn’t connect to external services.
Now to talk about AIs role.
This project was an inflection point on AI for me. I’ve used AI lots in the past, in 2024 it was good at line completion and some python functions. By 2025 it could give you alright code in many languages and frameworks, becoming handing in going from 0 to 100 in a particular subject in an hour. But I always still had to do the heavy lifting, I had to understand every piece of code, and make all the low level code decision because AI couldn’t be trusted.
I tried used cursor in December 2025 when I was creating my personal website, and it just still wasn’t good enough. AI coding agents really were more of a hindrance, as they just kept breaking my code and trying to give them direction made it break even more. But now by mid 2026, its impressive.
I didn’t write a word of code. I gave it my architectural design that was roughly in my head. I told it what frameworks and libraries to use, what external APIs to rely on. And it created it. Not perfect but with that little input of my plan, it gave me the 80%. And the cherry on top was the last 5 days of it was telling AI what I wanted changing and fixing … and it basically always did it, first try, and added tests so it wouldn’t happen again. This is really night and day.
I listened to a podcast this year where Mitchell Hashimoto roughly made the point, AI will now automatically do all those annoying little tasks that always blocked you from your project so you can focus on what you find interesting. I think back to all the times I wanted to create something but having to learn how a framework, library or service worked would just stop me because it was just too much work. Not anymore. I feel a new lease on life now to start rattling off some of these small tools I always wanted to crate.
I would also like to note, I did not use Claude efficiently when developing this. I kept all the prompts in one single context window that kept getting compacted. Thats to say, without me even trying to do some sort of efficient AI usage, it really did the job so well. So do I believe now that software engineering is dead?
Yeah to some extent. I’m sure it still struggles in much larger and complex systems and codebases. But it is already improving.
At the end of the day, businesses want to provide better service / products to customers at a lower cost. Thats the only metric they care about. If a customer can go somewhere else to get the same things for cheaper, the customer eventually will, and so the businesses will always follow the markets best practices for reducing costs. What that looks like in reality I think will be situations like, where there would previously be 8 developers to manage a website, that turns into one developer and an AI. Its just simply cheaper.