Alex Holcombe

Opensource code and libraries to fall behind as Anthropic seems to be giving only large corporations access to Mythos, including to find their software's security vulnerabilities. https://www.nytimes.com/2026/04/15/opinion/mythos-open-souce-internet.html

Google's gotten a lot more aggressive in their steps to kill the internet - now on my phone from Google.com webpage when I search for something, I only get AI results until I dismiss the AI window.

"I prefer the publish-review-curate model over the traditional submit-get rejected-submit somewhere else-get rejected-submit somewhere else model."

Found myself writing the above when a journal editor asked me to weigh in on accept/reject for a manuscript I reviewed.

Stay tuned as additional advocacy is in the works for the publish-review-curate model! #scholarlyPublishing

"Improving Acknowledgments Sections To Better Credit Research Contributors" v2 is now available on MetaArXiv https://osf.io/preprints/metaarxiv/t2uwe Took 2.5 weeks for MetaArXiv to approve the a new version, which I assume will get longer as more and more people use AI to do metascientific analyses and overwhelm the moderators.

I was interviewed by Oakleigh Wilson of @sortee@ecoevo.social about why our scholarly publishing system is the way it is and how to fix it. Thanks to Oak for doing a LOT of "editing for clarity"! 😀 https://sortee.org/blog/2026/03/03/2026_ecr_asks_alex_holscombe/ #scholarlyPublishing

Because of the increase in submissions due to AI, the peer review system will collapse. The AI flood is happening at different rates for different fields and subtitles, so you might be spared so far.
Other than falling back on old boys network, prestigious institutional marker type stuff, plus AI doing peer review, I dont know what will replace the current system. Even getting your preprint on a preprint server will be time-consuming and unreliable, depending on who you are, I think.

I want students to see examples of how shoddy research gets hyped and trace it back to the underlying work. For an undergraduate seminar, looking for suggestions for take-downs of popular psychology claims, based on critiquing the underlying flawed research. Below are the suitable ones I have so far.
"Compared to what?" is one theme I have; it's my phrase for highlighting that one always needs to think about the control/comparison condition and the possibilty of confounds.

Think of all the money research funders have poured into the maw of the Big Five publishers, paying for APCs instead of funding diamond #openaccess infrastructure, what a shame! https://scholarlykitchen.sspnet.org/2026/02/18/guest-post-diamond-open-access-needs-institutions-not-heroes/ "Open access needs institutions not heroes" by Curt Rice.

The volume of AI-enabled decent scientific work will overwhelm scientific journals and reviewers. Good post about this, except it ends on a strange note, that Wiley and Elsevier AI tools might save us, which is both unlikely and would give even more power to them.
https://open.substack.com/pub/thebsdetector/p/the-coming-apocalypse-for-scientific

"At a time like this, scorching irony, not convincing argument, is needed."

- from Frederick Douglass' 1852 speech, "What to the Slave Is the Fourth of July?"

Happy Birthday, Frederick Douglass!

"An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into accepting its changes into a mainstream python library." Pubpeer, journals are next!
https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/ via @UlrikeHahn@fediscience.org

#icanhazpdf https://doi.org/10.1212/WNL.0000000000214677 Would like to have this article! Possibly good example for my class of a new observational article treated uncritically by the press as causal (https://www.theguardian.com/society/2026/feb/11/reading-writing-lower-dementia-risk-study-finds)

A document created by the Heritage Foundation called “Wikipedia Editor Targeting" has surfaced.
Musk also attacked Wikipedia.
The Wikimedia Foundation says it will likely roll out features previously used to protect editors in authoritarian countries more widely. https://www.404media.co/wikipedia-prepares-for-increase-in-threats-to-us-editors-from-musk-and-his-allies/

We are not ready for the flood of AI-written fraudulent papers that will be increasingly difficult to distinguish from work that actually took place.

We need data chain-of-custody tools. Anybody using anything like that?