Hacker News new | ask | show | jobs
by girvo 4709 days ago
For what it's worth, I'm working on a Mint competitor of sorts (that takes advantage of Machine Learning to automatically help you save. It will be based in Australia, not the US, and the basic app will be released as open source for personal self hosting.
5 comments

I don't think any country is safe at this point. TNO (trust no one) is the only solution. Your cloud provider should have no ability to hand over your data because they can't decrypt it themselves. For example, Lastpass has an architecture where the passwords are encrypted and decrypted on the client, the server never sees anything but pseudorandom noise, and you can audit their browser addon to verify this. You can, with careful design, build many - if not most - cloud services in this way.
That is exactly what we are trying to do. The problem is that is somewhat at odds with machine learning in practice, but I have some ideas in the space.
Just an FYI, a Mint competitor, Wesabe, went out of business some time ago. When they did, they open sourced their software. Not sure of the state it's in. Maybe you can find some good things in their bank interface/scraping code...

https://github.com/wesabe/pfc

Brilliant! Thanks for that :)
After reading rlvesco7's comment, I immediately deleted my Mint account. It always made me uncomfortable to have all my financial data sitting in the cloud. I stopped actively using it a few months ago.

An open source service sounds interesting, but I don't think I'll ever be willing to post all my financial data to a web service again. It would be great to have a locally installed application that could keep track of all those accounts. Having some algorithms run to help me save would be great, but it would take some demonstrated assurances to get me to provide even anonymous data for the machine learning process.

This is interesting. Machine learning requires data. Will the app be sending back up anonymized data that then gets used to help the app make better decisions? Or how else will you make your app smart like Mint? Can't wait to see what you're doing.
The former is what we are doing, but coupling it with some basic statistical financial methods (and some stuff I've come up with myself!) that you can rely on if you'd rather not send the information. That's also what the cheapest plan for our hosted version relies on.

The entire premise is personal finance software that learns your habits to make it easier to use :)

That sounds good. I'd be interested in seeing the open source version