I really like this - but I'd rather just pay $200 once than pay a subscription. If an issue is maintaining the models, you should make that a separate line item

I really like this - but I'd rather just pay $200 once than pay a subscription. If an issue is maintaining the models, you should make that a separate line item where you can pay as a subscription for updates or pay for the model on a yearly/monthly basis as necessary (some people might be OK with the level of OCR at any given moment). What would be really interesting is if you could have some hardware that hooks into the HDMI port and with every flicker of the screen record and OCR everything, and add some diffing and compression and some really cool categorization algorithms and you could have something amazing. --- Obviously this is an MVP but I imagine the perfected version of this would be able to see the difference between screenshots and use some sort of strongly-connected-component like algorithm and figure out which things are "browser" which things are "browser--hacker-news", and so forth. Create a great UI around this idea and even better UX and you have something interesting. Hell, miniaturize some camera and hook it onto glasses and you have that one Black Mirror episode.