The "Brainrot Analysis" Project

[dennyabrain]

Overview

Can Feluda be used to study my Instagram Reels consumption pattern?

Smartphones show the number of hours you spend on an app and this screen time is shown as a scary number to be worried about. I think that all the hours I spend on Instagram Reels are not wasted. Some are spent looking for useful and informative content. Some are spent keeping up with culture and some for the guilty pleasure content. Its not the hours I am guilty about, its the absolute lack of control on what kind of content I spend time consuming.

Could we use automated web scraping and feluda to

1. Categorize the content I consume and produce a more granular understanding of what I am spending my time on Instagram on
2. Create a better way to consume Instagram reels where I have more agency over what I choose to watch in a day.

Reference

• This youtube video that blends storytelling and a playful ML experiment
• Tattle's study on Images collected and clustered from Whatsapp groups

[aatmanvaidya]

a nice visualisation of how the algorithm works - https://whatsthealgorithm.com/

[dennyabrain]

Upcoming fosdem has a talk on similar lines. It is limited to fediverse and hence they don't have the same starting point as ours. From the description it looks like they use llama to generate embeddings and make custom algorithms to generate feeds. That part of the talk could be relevant to our work.