James Zou

AI

Retest of 50 LLM findings classifies 39 as expired or weakened

Across 20 models, 22 findings were classified as expired and 17 as weakened by September 2026.

EM

CT

XS

12 Sources, 4d ago, updated 4d ago

AI

Posts discuss switching AI-agent harnesses mid-task

sanjana__z calls cascade harnesses recent work; james_y_zou claims switching can improve performance while cutting cost and token use.

JZ

YK

KZ

6 Sources, 19d ago, updated 19d ago

AI

Collaborative Intelligence benchmark aims to study when and how AI models make good collaborators

A researcher says their lab is interested in what future AI agent collaboration will need to be useful and safe.

JZ

SW

AP

11 Sources, 19d ago, updated 19d ago

AI

Stanford Medicine researchers turn scientific papers into AI agents

A researcher announced Paper2Agent’s publication in Nature on September 16, describing it as a way to turn papers into “virtual authors” that can answer questions and apply their methods.

JZ

NP

3 Sources, 40d ago, updated 40d ago

This is figure 1, which gives an overview of Paper2Agent.

AI

Physics of Agents Paper Models LLM Agent Dynamics

Stanford researchers model how LLM agents update opinions through statistical physics laws.

SG

RZ

JZ

14 Sources, 81d ago, first seen 81d ago

Recent posts

Retest of 50 LLM findings classifies 39 as expired or weakened

Posts discuss switching AI-agent harnesses mid-task

Collaborative Intelligence benchmark aims to study when and how AI models make good collaborators

Stanford Medicine researchers turn scientific papers into AI agents

Physics of Agents Paper Models LLM Agent Dynamics

James Zou

Recent posts

Retest of 50 LLM findings classifies 39 as expired or weakened

Posts discuss switching AI-agent harnesses mid-task

Collaborative Intelligence benchmark aims to study when and how AI models make good collaborators

Stanford Medicine researchers turn scientific papers into AI agents

Physics of Agents Paper Models LLM Agent Dynamics

Elsewhere on the web