A future improvement I am looking into is adding unit tests for tax calculation. If the project gets popular enough I would probably hire an accountant to verify everything.
When using a well-known site like TurboTax, while I'm totally appalled by their business practices, I have at least an assumption that I can do my taxes half decent (and not too much more than that either). It will be really hard for an open source project to gain enough momentum for me to have the same confidence. For a lot of other open source projects I can just try, and see what happens. But I'm not going to try and see what happens if I'm ever getting audited. And I'm using (begrudgingly) TurboTax exactly because I know nothing about tax code, so the unit tests (yet to be written) would not tell me much. I find this project very sympathetic, though, but I'll have to pass.
That's absolutely fair! I think finding a solution to this will be key to any kind of mainstream adoption
A practical self-test would be to enter up your last years tax return as is and compare the assessment to what you actual assessment was.
[Having proper unit tests as specs is a much cooler answer.]
Various tax law changes could make that inaccurate.
I think he means comparing output of the app for 2019 with what you filed for 2019. Not using the app for 2020 with your filings from 2019.