New paper: Emerging trends in AI agent evaluation

Understanding the methods, questions, and ecosystem for evaluation

Oct 19 • Sayash Kapoor and Arvind Narayanan

138

19

19

Steering AI Beyond Automation

A Research Agenda and Plan for Action

Oct 14 • Arvind Narayanan

198

38

30

A big-tent or small-tent AI safety movement?

The unstated disagreement that underpins safety debates

Oct 1 • Arvind Narayanan and Sayash Kapoor

215

91

44

September 2026

AI existential risk probabilities are (still) too unreliable to inform policy

How speculation gets laundered through pseudo-quantification

Sep 28 • Arvind Narayanan and Sayash Kapoor

169

31

26

The AI-as-Normal-Technology view of loss-of-control incidents

A middle ground between the cybersecurity and AI safety communities

Sep 14 • Sayash Kapoor and Arvind Narayanan

306

58

54

August 2026

AI agents can't yet do open-ended AI research

Early evidence from two case studies

Aug 5 • Sayash Kapoor and Arvind Narayanan

202

27

26

July 2026

What will be left for us to work on?

My keynote at ICML 2026

Jul 13 • Arvind Narayanan

357

30

64

Up the Stack: How AI’s Escape From the Commodity Trap Risks Enterprise Lock-in

Critics and boosters are both looking in the wrong place

Jul 9 • Arvind Narayanan

175

23

31

June 2026

Why AI hasn’t replaced software engineers, and won’t

Coding agents as normal technology

Jun 11 • Arvind Narayanan and Sayash Kapoor

621

92

114

May 2026

Did Google’s AI agents really build an operating system for $916?

The importance of independent evaluation

May 22 • Sayash Kapoor and Arvind Narayanan

147

25

11

Do AI Risks Require Extraordinary Government Intervention?

Let’s not skip the hard work of AI governance

May 21 • Sayash Kapoor and Arvind Narayanan

88

9

19

April 2026

Open-world evaluations for measuring frontier AI capabilities

Introducing CRUX, a new project for evaluating AI on long, messy tasks

Apr 16 • Sayash Kapoor and Arvind Narayanan

101

8

14