AI NEWSROOMBY CLASSWISE.IO

Saturday, November 21, 2026

OpenAI agents complete nearly half of research

Plus: Amazon and Google take large Anthropic stakes at IPO; Neural networks realize Wess-Zumino-Witten models

Welcome to a Saturday edition where the machines are quietly doing nearly half the work.

RESEARCH

馃敩 OpenAI agents complete nearly half of research tasks

OpenAI agents complete nearly half of research tasks
Image source: openai.com

OpenAI's coding agents completed 23% of two-to-four-hour AI research tasks without human help in March, rising to 46% in October. As of September 2026, the API-equivalent value of agentic AI use in OpenAI's research organization was about twice its non-equity compensation.

The details:

  • The median researcher at OpenAI used around $2,400 USD per day on agentic AI tools at API list prices.
  • Agent workdays per researcher at OpenAI rose from 0.5 to 5.7, an elevenfold increase.
  • The median respondent to an OpenAI survey would give up 40% of team headcount to restore frontier coding agents after a three-month loss.

First reported by openai.com, 68h ago 路 openai.com 路 aimagazine.com

IPO

馃搱 Amazon and Google take large Anthropic stakes at IPO

Amazon and Google take large Anthropic stakes at IPO
Image source: techcrunch.com

Amazon will own 14.7% of Anthropic after the IPO. Google and Nvidia each agreed to buy $10 billion worth of non-voting stock in the IPO.

The details:

  • Amazon's stake could reach 23.8% if its convertible notes are fully converted.
  • Anthropic has committed to spend at least $110 billion with Amazon between May 2026 and April 2036.
  • Anthropic's IPO was priced at $83 per share.

First reported by techcrunch.com, 78h ago 路 techcrunch.com 路 fortune.com 路 finance.yahoo.com

RESEARCH

馃 Neural networks realize Wess-Zumino-Witten models

The paper develops neural realizations of positive-level Wess-Zumino-Witten models using bosonic Wakimoto and fermionic Fourier-Berezin fields. Level-rank duality connects the charged sectors of the fermionic construction.

The details:

  • Explicit neural observables recover conformal block data in the fermionic construction.
  • The neural construction enables the transfer of information about unknown correlators between paired theories.
  • The paper constructs the Virasoro algebra from a neural free-field architecture.

First reported by arxiv.org, 95h ago 路 arxiv.org 路 arxiv.org

馃摪 Everything else in AI today

Researched and written by AI Newsroom, by Classwise.io

See how AI Newsroom is researched and checked

Get tomorrow's issue in your inbox.

Free, daily, 4:57 AM ET. Five minutes and you know what matters in AI.

Join for free