The essay got mentioned in Wired today:
https://www.wired.com/story/wikipedias-existential-threats-have-never-been-greater/

On Fri, Jan 16, 2026, 1:35 a.m. Erik Moeller <[email protected]> wrote:

> On Fri, Jan 9, 2026 at 3:41 PM Christophe Henner
> <[email protected]> wrote:
> > Here it is: https://meta.wikimedia.org/wiki/User:Schiste/what-now
>
> Thanks for the thoughtful essay, Christophe, LLM-assisted or not :).
>
> You hint at improvements to the reader experience, and I think there's
> a ton of work in that category that could make Wikipedia truly a joy
> to use, without even wading into the contentious terrain of AI.
>
> Wikidata already contains a huge wealth of information that could be
> made more visible in Wikipedia. The external identifiers alone for any
> topic could give you a better "explore" page than Google's cluttered
> search result pages.
>
> A huge reason readers may prefer AI summaries, even if inaccurate, is
> to get to an answer more quickly (the same reason Wikipedia itself
> outperformed other information sources even when its quality was still
> very uneven.) But Wikimedia has a lot going for it: no ads, no weird
> AI hallucinations, no intentional clutter. Readers _want_ to like it.
>
> Consider the humble disambiguation page. Often, when readers look for
> something not covered by existing wayfinders, the best option is to
> scroll your way through a multi-page sectioned list of uses of the
> term. Could a data-informed design get the reader to the right
> selection much faster?
>
> Maps, media, and interactive learning materials have often been
> discussed on this list. Giving editors more powerful cross-media
> authoring tools will likely yield significant dividends.
>
> Regarding AI, I personally believe it can be the basis of powerful new
> research and authoring experiences. However, opinions on this topic
> are extremely polarized. Many reject generative AI, because they
> believe it cannot deliver on anything it promises, because of the
> harms it is causing (and there are undeniably many), or both.
>
> Wikimedia will likely be well-advised to keep the technology somewhat
> at arm's length, which I think is consistent with your essay:
> Wikimedia wikis as a trusted foundation.
>
> For folks who _are_ interested in experimenting with it as an
> authoring tool, I took the occasion of the 25th anniversary as a kick
> in the butt to get a small prototype up and running of an agentically
> edited wiki. It's currently at https://agpedia.org/ and it is a _very_
> quick and dirty hack [1], but I'm going to keep at it to push my own
> thinking forward. :)
>
> The idea here is that advanced AI tools can, in principle, mediate
> authoring. "Research this topic, recommend some citations, check out
> this PDF, now create a first draft, etc." I believe that this needs to
> happen with humans firmly in the loop and accountable for edits they
> make with the help of these tools, as opposed to the Grokipedia
> approach, which is basically an ideologically poisoned accountability
> sink.
>
> I doubt Wikimedia will go down this road anytime soon (or ever), but
> if anyone reading this is interested in collaborating on this or
> similar ideas, feel free to give me a ping!
>
> Warmly,
> Erik
>
> [1] https://github.com/permacommons/agpwiki (again: prototype!)
> _______________________________________________
> Wikimedia-l mailing list -- [email protected], guidelines
> at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and
> https://meta.wikimedia.org/wiki/Wikimedia-l
> Public archives at
> https://lists.wikimedia.org/hyperkitty/list/[email protected]/message/JTX7QM3MD76G4EY6FLDWMCAMJBR7XVCT/
> To unsubscribe send an email to [email protected]
_______________________________________________
Wikimedia-l mailing list -- [email protected], guidelines at: 
https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and 
https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at 
https://lists.wikimedia.org/hyperkitty/list/[email protected]/message/H6DWZDJTGW7FESZTBHXK7OIRQP5OCVSF/
To unsubscribe send an email to [email protected]

Reply via email to