priya

Stephan Guyenet's early deep dive on GLP-1 agonists causing 18-20% body weight loss - tracing the history from WWI explosives to modern neuroscience. Published before the Ozempic craze went mainstream.

“This was published before Ozempic was a household name. Guyenet saw the revolution coming and explained the science clearly. If you read this early, you understood why GLP-1s were going to be massive.”

Morgan Housel's original essay (before the book) on why financial decisions are more about behavior and psychology than financial knowledge — and why being 'smart' about money is mostly about being consistent.

“The book is good, but this original essay is better — tighter and more direct. The core insight that finance is mostly a soft skill is underappreciated by everyone who approaches it as math.”

Simon Willison's exhaustive year-in-review of every major LLM development in 2023. Dense with links and actual signal amid massive hype.

“Best factual record of what actually happened vs. what the press said happened. Simon is relentlessly empirical — no hype, just what he tested and observed.”

Saloni Dattani's roundup of the most important medical advances - from GLP-1 agonists for addiction to gene therapy breakthroughs. Rigorous, well-sourced, no hype.

“Saloni is a founding editor at Works in Progress and researcher at Our World in Data. Her science writing is the gold standard - technically rigorous but accessible. This roundup is a perfect example.”

Lilian Weng (OpenAI) on the underappreciated role of data quality in model performance - why more data isn't always better and how to think about data curation.

“Lilian's posts are basically free graduate courses. This one challenges the 'just throw more data at it' mindset with real evidence.”

Quanta's explainer on elliptic curve cryptography — one of the most elegant and practically important areas of modern math, now securing almost all internet traffic.

“Quanta is doing something rare: making actual math legible to readers without dumbing it down. This piece makes ECC genuinely intuitive.”

A structural analysis of why culture wars are a permanent feature of pluralistic societies - not a temporary dysfunction but a consequence of genuine value conflicts.

“Refreshingly non-partisan. Instead of blaming one side, it explains the game theory of why both sides keep escalating. Changed how I consume political media.”

Zvi Mowshowitz's analysis of whether RSPs from Anthropic, OpenAI, and DeepMind are real commitments or PR moves - with specific criteria for evaluation.

“Zvi does what nobody else does - reads the actual policy documents line by line and checks them against reality. The scorecard approach is exactly the accountability these companies need.”

Gwern's comprehensive exploration of the idea that intelligence emerges from scale alone — and what that implies for the future of AI research. Written in 2020, reads like prophecy now.

“Gwern was right when almost nobody else was. Reading this after GPT-4 exists is eerie — the predictions hold up better than most academic papers from the same period.”

Paper review of a phenomenon where distributed systems get stuck in a degraded state that persists even after the trigger is removed. Real examples from AWS, Azure, and Google.

“This failure mode is terrifying because the system looks healthy by most metrics while actually being stuck. I've seen this pattern in production twice and didn't have a name for it.”

Anthropic identifies millions of interpretable features inside Claude

“If we can understand what is happening inside these models, we can actually verify alignment claims”

Noah Smith's data-heavy argument that immigration is one of the few economic policies with overwhelming evidence in its favor - covering wages, innovation, and fiscal impact.

“Noah does what most opinion writers won't - actually looks at the data. The section on immigrant entrepreneurship rates is striking.”

Practical guide to building evaluation pipelines for LLM apps

“Most teams skip evals and regret it. This is the most practical guide I have seen.”

Solar is growing faster than any energy source in history

“Solar cost curves continue to beat every projection - this has massive implications”

Google DeepMind open-sources massive robotics dataset

“Open datasets for robotics could do what ImageNet did for computer vision”

Breakthrough in combinatorial mathematics

“Expander graphs underpin everything from error-correcting codes to cryptocurrency”