2 comments

[ 0.24 ms ] story [ 11.0 ms ] thread
[flagged]
Interesting ... as much as I wouldn't trust my data with online AI - I'll give your project credit for addressing both mobile and desktop in application - while noting your site is robust - meaning the introduction page works with my old browser - so many show hn projects do not.

Using AI I see some short comings and I wondered how you've engineered a solution to deal with certain issues.

a) docket isn't the clearest ... the printer is blurring a small section x characters from the left or right that obscures the full transaction or receipt number.

b) docket has not included a proper description and abbreviation used isn't the clearest. Worse it's a fuel docket and the muppets who set up the store register decided to save ink by not printing the litres and or cost per litre.

c) setting a percentage of private use for certain items. Things like power are easy, but one off purchases differ greatly depending on their use.

d) correlation issues with the bank statements of the company name on bank statement verse real world name, which the bank unfortunately uses both - possibly depending on what atm or eftpos was used in store.

With (d) this would be most likely overcome as the data set of a number of customers grew.