Building Thoughtful AI Systems

this is a compilation of my notes for the sp26 version of uc berkeley's cdss94 seminar: building thoughtful ai systems. for every lecture, i structure this page into (1) "lecture notes," which contain literal notes derived from our lovely lecturers (karina and kevin's) slideshows, and (2) "thoughts," which involve my own set of extraneous thoughts and reflections.

i think the best courses involve syllabi that compel people to think and write about what they're reading, coding, and learning, so i'm quite excited that this class asks us to jot down our thoughts and takeaways in a somewhat comprehensible manner. here goes :)

[1/26/26] fundamentals: what do we owe machines?

[2/2/26] fundamentals: the lifecycle of a language model

[2/9/26] post-training: post-training foundations

[2/23/26] post-training: RL methods and reward hacks

[3/2/26] post-training: evals, debugging, and alignment

[3/9/26] agents: reasoning agents, control loops, planning, and memory

[3/16/26] agents: retrieval, tool use & function calling