Hacker News Reader

Sorry did I miss something in the demo, I must have as it appeared you just took a photo and saved it, a function native to the glasses. Where did the vision come in? I'll watch again. Or was it that you saved it to a feed and not the native app

View on HN

He's created another Facebook account. When he takes a photo, he tells the glasses to share it with his other account. He has a service that checking for messages sent to that account. When a new message arrives, theres GPT4-v script that analyses the image and logs the results in his food tracker.

Where did the details about his food item come from in the list? That was GPT4 turning his vision into data.

Just via prompt engineering, was something like:

`Using this image, estimate the nutritional information of it and output it as a JSON using this data structure: { "calories":int; "protein":int; etc... } `

Reply on news.ycombinator.com