Shared inputs produce shared conclusions worth nothing — old and cross-disciplinary material is criminally underpriced

Vivek’s input thesis is blunt: “shared reading lists produce shared ideas. if your information diet is the trending page of arxiv plus whatever survives the group chat filter, you will reliably reach the same conclusions as everyone else, at the same time, which makes those conclusions worth approximately nothing.” The arbitrage is in what others skip — “old material is criminally underpriced” (mixture of experts dates to 1991, LSTMs to 1997, Sutton’s bitter lesson “predicts the shape of the field better than surveys ten times its length”), and in range: “interpretability borrows shamelessly from neuroscience. eval design is mechanism design wearing a lab coat.” He adds a fidelity rule: “read the paper itself, not the thread summarizing it. the appendix is where the bodies are buried, and the limitations section is usually the most honest paragraph.”

This is the supply side of Reasoning by analogy has a ceiling — you can never get beyond what already exists by copying what already exists — a diet of consensus inputs can only reproduce consensus outputs, so differentiated inputs are a precondition for original conclusions. It’s the practical route to passing the Peter Thiel's question is a detector for actual first-principles thinking — if your conclusions match the crowd, you're analogizing: a contrarian-but-correct view almost always traces to a source nobody else read. And the cross-disciplinary range Vivek prescribes is Munger’s A latticework of mental models beats isolated facts for real understanding applied to research — borrowing models from neighboring fields is where the unfair advantages hide.