How LLMs See Images, Audio, and More
A video is tokenised by first splitting it into individual image frames and audio sequences and then tokenising images and audios. Not surprising but interesting.
Bookmarks RSS feed
-
-
Learning to learn | K/L
aka avoiding the expert beginner trap.
-
WorkLife: How you can do more for others with Rutger Bregman
Conventional wisdom holds that if you are a good person, you will do good for others. The school of moral ambition says, you become a good person by doing good. A powerful reminder about how much we are in control of doing good deeds.
-
What are Forward Deployed Engineers, and why are they so in demand?
Now that is a role I’d love to be in, maybe a senior/principal FDE or something.
-
Untitled
Learning from others’ failures is the optimal way to learn - the 3D ringside seat method.
-
EP175: What is the SOLID Principle?
Lovely illustration of SOLID principles.
-
How Uber Eats Deduplicates Hundreds of Millions of Product Images
Relying on file hashes was my first guess too.
-
How LLMs See the World
-
Untitled
A great collection of examples of applying second order thinking in software development.
-
The State of React and the Community in 2025
A refreshing take on React’s history, evolution, and commercial entwining.