researching user agency in computing @MIT_CSAIL / building @TwemexApp / 🇯🇵🇺🇸
Thanks @magrawala, really enjoyed your post! The point about predictive conceptual models cuts to the essence of the problem
Great post. I think the point about “hard to read edits without a diff” can be generalized too: plaintext is a suboptimal data viz for so many LLM outputs. We need diff views, weather widgets, mini maps, 2x2 charts, …
Exciting post! Curious how you think about the challenge of teaching GPT a new language when there’s not much example code in the training set