VIDEO SCRAPING: EXTRACTING JSON DATA FROM A 35 SECOND SCREEN CAPTURE FOR LESS THAN 1/10TH OF A CENT

Summary from the TLDR newsletter

It's possible to record a screen capture of a browsing session and then use Google Gemini to extract numbers from the video. While these language models shouldn't be trusted not to make mistakes, it appears that the Gemini 1.5 Flash 002 model is sufficient enough to complete the task consistently. Extracting data from a 35-second video costs less than a tenth of a cent. Video scraping allows developers to have total control over what is exposed to the AI model. This article walks readers through how to do it.