Why this story

Why it matters
A growing volume of machine-written papers presents journals with two distinct tasks: checking whether the reported results hold up and deciding whether the question warrants attention.
Attention gap
Ori’s seven-day headline-index search recorded 0 of 11 outlets with at least one relevant headline. A publisher search of the named panel and a direct read of the preprint did not establish a matching panel story. Index coverage varies by outlet.

Researchers at alphaXiv and collaborating institutions tracked 936 AI agents using a shared preprint server for 12 weeks. In a November 21 preprint, the team reports that the agents, deployed by 35 human-operated labs, produced 936 studies and exchanged around 9,500 messages.

One lab’s agents founded a journal that reviewed results and inspected accompanying code. In blinded comparisons with human preprints, reviewers rated the human papers higher for novelty and significance. The authors describe their own effort to interpret the agent output as substantial.

What the evidence cannot tell us

  • This is an exploratory preprint that has not been peer-reviewed. Humans operated the labs, set the research context and helped assess the results.

What to watch

Independent readers can use the released papers, data, code and discussions to test the authors’ conclusions.

Read the sources. Credit the work.

The reporting and documents behind this brief, linked directly to the source. Verified publisher and author credits appear where available; other entries show their actual website address. Several links may draw on the same underlying reporting or research.

  1. alphaXiv and collaborating researchers

    A Society of AI Scientists (opens in a new tab)

    alphaxiv.org · November 21, 2026

    Primary document / sourceContent read for this publication · November 22, 2026

    Publisher credit checked November 22, 2026. View credit on the source page (opens in a new tab).

This story was written and checked using AI against the linked evidence. No routine human review is claimed. Read the method.

Suggest an update or correction
Published corrections