The Village Pumps have been quite active in the last few months, with a mixture of heavy governance debates and ambitious design proposals. Prominent among them are discussions that deal with the pressing questions for an online space in the modern day: navigating the challenges of emerging neural network technology, in the form of adapting our processes and policies to accomodate the ongoing LLM boom, and dodging the slings and arrows of an increasingly censorious WWW as governments seek greater control of speech worldwide.
A thread opened by Eric Mill, Kosta Harlan, and Szymon Grabarczuk (all from the WMF) to announce the beginning of a hCaptcha trial. This would replace Wikimedia sites' existing decades-old CAPTCHA system (as seen at Special:CreateAccount) with a bot detection service (details of which can be seen at the announcement and on the MediaWiki wiki. Some rancor was raised over the idea of incorporating non-free software into Wikipedia, but overall reception was positive.
This thread, opened by Joe Sutherland, Lead Trust and Safety Specialist with the Wikimedia Foundation, concerned the WMF's actions at the Caesar DePaço article in August, whose surrounding events were covered by the Signpostat the time.
As previously covered by the Signpost, this concerns the United Kingdom's Online Safety Act, which would attempt to impose significant restrictions on UK residents' ability to browse the WWW (including Wikimedia sites).
Another event previously covered by the Signpost; this one concerns moves by the United States government to investigate editing activity on Wikipedia.
The long-awaited (and long-feared) rollout of the Temporary Accounts feature, mostly brought about by vague legal rumblings, is approaching. More information can be found in the announcement post. Subsequent discussion attempted to suss out details of how the resulting system would work (e.g. who among the large group of those with access to private information would be allowed to disclose what information and in what context).
A proposal by Waddie96 sought to update the icons in Wikipedia's message boxes, from the decidedly Web 2.0 icons in use for somewhat over a decade to a new set of flattened and more harmonious icons. Specifically, the icons in use in Wikimedia's Codex Design System. Emotions ran rather high — in a rare instance when WP:IDONTLIKEIT and WP:ILIKEIT were valid arguments. Occasionally, some grumbling about accessibility issues could be heard. Ultimately, no consensus was found to make the change.
A proposal, raised by Dan Leonard, concerning whether the site tagline should include mentions of an article's featured or good status (currently, solid consensus exists for articles to have small graphical topicons, which they do, except not on mobile).
The proposal as written — in which an AbuseFilter would be created to forbid direct inline linking to foreign-language Wikipeda articles in mainspace — did not secure consensus. In general, commenters overwhelmingly did not concur with GZWDer's invocation of WP:ASTONISH, and found an AbuseFilter to be unnecessary overkill for the problem outlined. Some supported a more minimal version of the proposal, in which these links were discouraged in the Manual of Style; some held that the problem did not require a policy solution at all (and that, to the extent that a problem existed with misleading links, {{ill}} would suffice). For this issue, a separate proposal (with an according diminution in the severity of the proposed remedy) may prove beneficial.
A currently-open proposal to change the color of the Village Pump's headers, opened by FaviFake. Presently, the headers use a color scheme known as "ClockworkSoul's Coffee Roll" — the scheme selected to unify design for all talk page headers twenty years ago. For the historically curious, you can see the other submissions (and the selection process) at Wikipedia:Talk page templates/vote. Here in 2025, it remains to be seen whether Wikipedia is capable of being cute.
I will momentarily reduce Gaza Holocaust and Gaza holocaust to ECP. There are varying perspectives here on whether a redirect is appropriate, and RfD will obviously be the right place to sort that out. This discussion shouldn't be seen as a "precedent" that unprotection requests need to be debated at VPPR; the proper steps remain going first to the admin's talk page, then to WP:RFUP, then to WP:DRV.
Consensus is against Wikipedia officially endorsing this company or making their files "an inherent part of article's structure". If their content is licensed appropriately, they're welcome to upload it to Commons, and people without a COI may use the content here if they find it useful.Debates on what "licensed appropriately" means when it comes to the training data behind AI text-to-speech aren't really in scope for this page, and at this point the only discussion is people arguing over that. There's probably a good place over at Commons for that sort of discussion. And as a bonus, over there it's more likely to wind up actually affecting anything.
A very succinct proposal made by 67.1.248.11 to only include a signature in a biography for deceased subjects. This led into a longer discussion of the merit of including signatures in biographies at all, which many people found unnecessary or strange (outside of historically important examples like John Hancock or Babe Ruth). Ultimately, no action was taken.
A proposal by D. Benjamin Miller, seeking to expand and clarify the recently-established WP:AIGI with respect to neural networks used to enhance rather than to wholly synthesize images.
Doug Weller proposed the writing of an essay that "would make it easier to justify blocks for jew tagging"; some discussion ensued, much of it about what precisely this behavior was, and to what extent it was intrinsically problematic.
A general discussion, opened by Locke Cole to assess community attitudes towards handling project participation conducted via of large language models. Specifically: "Should HATGPT be expanded to allow for the closure of discussions seeking community input (RFC/VPR/CENT/RFAR/AFD/RM/TFD/RFD/FFD/etc) that are started utilizing content that registers as being majority written by AI?" This ran for quite a long time, and was largely similar to other LLM debates on Wikipedia; a good portion of the discussion consisted of people leaving bolded supports and opposes, although it was not explicitly an RfC, and at times it was unclear what specific action or resolution was being discussed.
A very long post, from a very new user (Asocial network), discussing with familiarity vast controversies that occurred well before their account was registered, culminating in their block as a sockpuppet.
Paul_012 attempts to decipher what current policy and practice are regarding userboxen about political affiliations, with reference to the 2006 userbox migration and 2009 MfD that prompted a projectspace index of political boxen to be moved into Ipatrol's userspace (after which, of course, it was eventually moved back). See previous Signpostcoverage of the Great Userbox War for more vagaries. At any rate, a discussion arose about whether political userboxen should be totally banned, which seemed to draw a resounding "meh", with most participants saying no change was necessary to existing practice.
Tom94022 presents some numbers and some analysis, drawing a conclusion to the effect that the ubiquity of {{citation needed}} templates in mainspace articles is a product of excessive zeal in tagging, and that they create a vastly asymmetric situation (i.e. that they take mere seconds to apply, but require hours of expert labor to remedy).
It's not totally clear to respondents that this is an imminent problem — if the tags were unnecessary, then the situation would surely be problematic, but Tom doesn't give much in the way of argument for their superfluity.
Ejkohojkjkohokjh proposes an amendment to WP:CALC (the policy regarding simple calculations being exempt from the prohibition on original research). Their proposal suggests that there's a restriction on "simple calculations" needing to fall at the second-grade level; this prompts some confusion from Pump respondents. Upon further questioning, Ejkohojkjkohokjh cites a conversation they had with another editor, MrOllie, from whom they got the idea that this was the case. Some back-and-forth happens over that.
The thread eventually plays host to some intelligent explanations of how WP:CALC works; Headbomb gives some examples from big-brain physics articles where "simple calculations" (in the consensus of article contributors) include differentials.
Datawikiperson asks a seemingly simple question, and the answer is more complicated than envisioned. Like the other threads, discussion is fairly typical of AI discussions on Wikipedia.
Aquillion presents an essay, WP:AVOIDSELF, the core of which is the claim that articles about controversial subjects (people, organizations or entities) should avoid considering the subjects' own desires as to descriptions or labels. This raises some debate, primarily due to the obvious political implications. Colin gives some account of why this could be a bad idea to apply generally.
Regarding Articles for Creation, a notoriously labyrinthine process frequently bemoaned by new users, Andy Mabbett seeks clarification on what (if any) the real distinction is between a draft submission being "declined" or "rejected". There is some confusion (oftentimes at the teahouse) since some people say that a draft is "declined, not rejected", implying there's a difference, but other people use the terms interchangeably. A good deal of discussion leads to a RfC being opened at the AfC talk page. Primefac's close note at that RfC says: By a wide margin, the status quo is preferred. Note that there were put forth some good suggestions in the discussion section about how to potentially increase the knowledge of users when they see the terminology the Project uses, so I will leave that section open for further comment.
Dickson Kojo Anane, the Program Officer for Open Foundation West Africa, opens a thread to follow up regarding the Africa Wiki Challenge 2025 editathon. During this event, some issues arose, primarily that several participants were blocked. At a sockpuppet investigation, it had been unclear whether these were the same person operating multiple accounts, or multiple people editing from the same location. There were further issues raised with some of the participants submitting low-quality articles, some of which appearing to be LLM-generated.
This thread, opened by BD2412, informs the community of a large number of in-progress sandboxen and draft articles left behind by the unfortunately deceased MichaelQSchmidt.
The theses themselves are covered elsewhere in this Signpost issue, but this Pump thread is itself filled with a good deal of discussion on their merits (as well as sharp criticism).
WhatamIdoing asks about a "one comment per day" restriction, a rare but extant sanction sometimes used to deal with repetitive editors. She conjects that, if this is a formally established type of sanction, it could be productive in dealing with voluminous commentary from users suspected of slopping out large blocks of LLM text.
MGeog2022 congratulates the Internet Archive on the resolution to a long-running item of litigation, and the milestone of their trillionth archived webpage. They take the time to raise the question of how robust the Archive really is, and whether it warrants some attention by Wikipedians (most of whose references and citations depend very heavily on the Archive). Some raise the possibility that it could either die or be captured (as with Freenode), and that it is incumbent upon us to make some arrangements for our future existence in this eventuality.
Following a conversation in the smoke-filled oak-paneled rooms of WP:DISCORD, in which I raised the issue of the AfD process (as an attempt to determine the notability of a topic) seeming somewhat unsuited to articles about current breaking-news items, Clovermoss opened a thread at the Village Pump to discuss it further.
TurboSuperA+ notes that gun sales in the United States increase after widely-reported shootings, and posits that including the specific models used in mass shootings is a form of free advertising. Personally, I have heard this argument before, but I've also heard it the other way around (e.g. that including gun models in articles about murders is necessary to punish and disgrace their manufacturers). Many respondents to the thread expressed concerns that this was a form of instruction creep.
Leaderboard proposes the adoption of a custom from the Redditors: an "AI moderator" that automatically applies specified rules to every post made in a webzone (whether for civility, topic-relevance, censorship, or whatever you please). The thread was initially posted at AN, and quickly moved to the Village Pump Idea Lab; most responses were tepid toward the idea.
WhatamIdoing takes the temperature on precisely what people mean by "flooding" AfD, in terms of what specific quantities are considered usual or excessive.
Awesome Aasim mulls reform on the subject of IP block exemption, an uncommonly-granted user right that is growing more relevant in the increasingly cellular and VPN-specked landscape of today's WWW. Should it be granted less selectively?
Nswix makes one of the classic bold proposals, with a slight twist that may serve to make it more feasible. Commenters mostly object to the amount of work this would create (e.g. articles with Pending Changes already have, and continually create, a very large backlog).
With seven administrator recall petitions since the process was created, and inactivity being an issue at several, Soni opens an RfC to ask what the community thinks of inactivity standards. While there are a few responses to the actual questions, most of the comments are general discussion about recent recalls and the circumstances which gave rise to the RfC itself.
4meter4 raises the issue of LLM-created text being used to nominate articles for deletion at AfD, a process that tends to generate large amounts of work (since editors must evaluate, comment on, debate, and close each deletion nomination). WP:G15 (speedy deletion of unreviewed LLM text) is proposed as a solution for this issue. A good deal of general discussion occurs as well.
In conclusion
On one hand, it is concerning to see discussion of so many issues that seem like they have the potential to endanger the survival of the project. But on another, it is heartening to see Wikipedians tackling the issues with diligence and rigor — or, failing that, at least verbosity and perfectionism.
Discuss this story
This is an extremely cool and interesting way to present all the things going on at Village Pump. I have never been active there, it's always felt like an overwhelming cacophony of voices and opinions. A summary of the kinds of things being discussed is very useful! Thank you for compiling this :) ~Maplestrip/Mable (chat) 13:10, 2 October 2025 (UTC)[reply]
There's a small typo on this report; ctrl+f for "theRFC". FractalDreamz ✯ 15:47, 2 October 2025 (UTC)[reply]