A collection of texts that I found to be interesting. These pieces may not necessarily represent my views, but I found them fascinating enough to bookmark and revisit.
by OpenAI
The story behind GPT-5.5’s bizarre “never talk about goblins” system prompt and the reward signal rabbit hole that explains how it got there.
by Scott Alexander
An allegory on the modern debate over Artificial Intelligence, questioning if it's a true emerging intelligence or merely a sophisticated mimic, starring God, Iblis, and Dwarkesh Patel
by Apple Machine Learning Research
Compares difference in capability between LRM's and LLM's in various tasks, and asks a critical question: Are these models capable of generalizable reasoning, or are they leveraging different forms of pattern matching?
by Jason Fried
Reflects on overcoming his hotheadedness tendencies by learning to pause and think for five minutes before dismissing ideas, inspired by Richard Saul Wurman's advice, emphasizing the fragility and potential of new ideas
Discusses V-JEPA, a self-supervised learning model for video understanding, highlighting its ability to predict actions and model complex video dynamics effectively.