PERSON DIRECTORY
Lisha Li
Lisha Li appears in 3 indexed conversations across The a16z Show. This directory brings every appearance, source, TL;DR, digest, and transcript into one searchable feed.
Can Open Models Keep Up With OpenAI and Anthropic? | Nathan Lambert
Lambert estimates distillation may give Chinese labs a few months, but its value in RL-heavy post-training remains unclear.Closed labs integrate inference and lead coding, where revenue has gone from zero to $100 billion, while open-model economics fragment builders, inference firms and post-training companies.Open models may still win buyers seeking owned weights and the cheapest adequate model.
Inside OpenAI’s Breakthroughs in Mathematical Reasoning
Lisha LiMehtaab SawhneyMark Sellke
Astra determined the asymptotic Cohn–Elkies linear-programming bound for high-dimensional sphere packing, constructing an attaining function and proving no better function exists within the framework.Its edge is execution: correct bets, technical precision, and clean restarts produced a roughly 15-page non-sofic-group proof, while task-oriented stopping leaves supervision and long-horizon execution as risks to monitor.
What Today’s Best Models Still Can’t Do in Math
Litt’s favorite autonomous signal is the mid-May solution to the Irish unit-distance problem, whose classical 1960s techniques helped mathematicians find counterexamples elsewhere, including the sum-product conjecture over the real numbers.Models remain weak at theory-building and long-proof verification, while slot-machine papers risk mode collapse, making human-led research diversity and the accumulation of understanding important watchpoints as capabilities advance.


