Reading
A living shelf of the books, papers, and articles shaping how I think about machine learning, systems, and research taste. Some left a permanent mark, some are in progress, and some are queued because they seem worth wrestling with next.
Highlights
- BookDesigning Data-Intensive Applications
Martin Kleppmann
Still one of the clearest books for developing taste around system design tradeoffs.
systems · distributed systems
- PaperLearning Transferable Visual Models From Natural Language Supervision
Alec Radford et al.
A benchmark paper for understanding how large-scale weak supervision changes transfer learning.
computer vision · multimodal · representation learning
- ArticleThe Bitter Lesson
Rich Sutton
Short, sharp, and annoyingly durable whenever model progress starts to look surprising.
ml philosophy · scaling · research taste
Read6
- BookDesigning Data-Intensive Applications
Martin Kleppmann
Still one of the clearest books for developing taste around system design tradeoffs.
systems · distributed systems
- PaperLearning Transferable Visual Models From Natural Language Supervision
Alec Radford et al.
A benchmark paper for understanding how large-scale weak supervision changes transfer learning.
computer vision · multimodal · representation learning
- ArticleThe Bitter Lesson
Rich Sutton
Short, sharp, and annoyingly durable whenever model progress starts to look surprising.
ml philosophy · scaling · research taste
- PaperFlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
Tri Dao et al.
A strong example of how hardware awareness can create algorithmic leverage.
transformers · systems · gpu
- PaperAttention Is All You Need
Ashish Vaswani et al.
Worth revisiting for how many now-standard design decisions started as sharp simplifications.
transformers · nlp
- BookSystems Performance
Brendan Gregg
Useful for keeping systems intuition grounded in actual bottlenecks instead of folklore.
performance · observability · systems
Reading3
- BookUnderstanding Deep Learning
Simon J.D. Prince
Reading it as a way to tighten up intuition around the pieces that usually get hand-waved.
deep learning · theory
- PaperDeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
DeepSeek-AI et al.
Current focus for understanding where reinforcement learning is buying reasoning behavior.
llms · reasoning · rl
- ArticleThe Illustrated Transformer
Jay Alammar
A useful companion when I want a quick mental reset before reading more technical variants.
transformers · explainers
To Read3
- BookThinking in Systems
Donella H. Meadows
On deck because I want better language for feedback loops and second-order effects.
systems thinking · decision-making
- PaperThe Perils of Reward Model Overoptimization
Leo Gao et al.
Queued up to sharpen my intuition for reward hacking and offline evaluation gaps.
alignment · rlhf · evaluation
- ArticleThe Chips That Power AI
IEEE Spectrum
A lighter-format read to complement the lower-level GPU architecture posts on the site.
hardware · accelerators