I'm starting to see some reviews of Sunshine Hills Reality Edit (Tina Connolly), in a post-truth world with a Wikipedia-like entity. The book doesn't even have an article yet as far as I can see. Going to wait just a bit and see if a) it is notable and b) if more reviews appear. ☆ Bri (talk) 00:42, 26 September 2026 (UTC)reply
Sorry to add this at the 11th hour. Frankly the story turns my stomach, but it needs to be told calmly. It looks like the guy who supervised Christa Pike's botched execution supervised 3 other executions in Arizona in 2022 and likely told the executioners to look up the proper dose on Wikipedia. And the appropriate Wiki article provides (still!) that information (despite the MOS for medicine strongly saying that we shouldn't provide doses).
I know I don't write well when the story turns my stomach, so extensive copy editing for style would help, as would some independent fact checking. For some reason I can't bring myself to name and link the Wikipedia article where the dose is given. Something about "what if somebody else uses it?" Any help appreciated, but please keep it dignified and conservative and don't put it in the title or blurb. Smallbones(smalltalk)17:56, 8 October 2026 (UTC)reply
Yeah, the NPR story is perhaps the best story to base this on. I've got a lot of on-Wiki stuff sorta related to this. But it really doesn't make a lot of sense. Other than some good Wiki editors ignoring the Meds MOS, there's nothing I can say about them doing anything wrong or acting in bad faith. It's just very bizarre that a government official seems to have told the executioners working for him something like: "Hey guys, you want to kill this person right, just check Wikipedia". But no jokes allowed here, ok? @Bri, HaeB, Wil540 art, JPxG, and Czarking0: any ideas? Smallbones(smalltalk)21:54, 8 October 2026 (UTC)reply
Not anybody's fault here, but I don't love the passive voice in the NPR story: in the records when department employees are told to check Wikipedia. What records? Who says those records exist? Who wrote the records? How were they disclosed? Who told whom to check Wikipedia, and what were they supposed to check? Was it the only reference used for the lethal doses in execution protocols? Was Wikipedia even consulted about dosage, or something else (potency of aged chemicals, maybe)? It's just so thin – far less than a sentence of substance. ☆ Bri (talk) 22:30, 8 October 2026 (UTC)reply
There is at least one indisputable fact here, that Frank Strada was in charge of 3 botched executions in Arizona and then in Tennessee he was in charge of Christa Pike's super-botched execution, the buck stops with him if ultimately he (or the states) cannot produce where they got the dosage or what was the protocol for getting it. Judge Duncan's expertise only extends to his investigation in Arizona and he just refers to "records". Are they printed protocols or emails or maybe even just an interview with an execution staff member. He doesn't say. His whole point in that short paragraph is that many records of protocols are intentionally being hidden (by the state or its employees) so that they can't be held accountable. So does that mean that we don't report what the media has said, or that we just report what we know from the media and from the record as reported in the Wikipedia database of records. I've added that I found "little or no" evidence showing that Wikipedia editors broke any rules. More evidence of course would be better than the little we've got. But I feel that publishing what we've got is a lot better than not publishing it. Smallbones(smalltalk)00:23, 9 October 2026 (UTC)reply
A lot of users have said that the news was very concerning If I was writing this piece, I would not say this. I would say something more along the lines of Several users such as Username 1 and Username 2 indicated heightened concern for the reliability of CBS. I would then reach out to Username1 and Username 2 for comment and find new information about how they think about the future of news.Czarking0 (talk) 20:09, 19 September 2026 (UTC)reply
As usual, we are preparing this regular survey on recent academic research about Wikipedia, doubling as the Wikimedia Research Newsletter (now in its sixteenth volume). Help is welcome to review or summarize the many interesting items listed here, as are suggestions of other new research papers that haven't been covered yet. Regards, HaeB (talk) 17:16, 7 October 2026 (UTC)reply
I do not believe that the results will be a surprise to most Wikimedians who have been curious about generative AI. It is however disheartening to see that the journalists and science communicators we may be trusting to provide high quality, verified information, do not seem to share our caution. This is not an opinion piece so authors should not express their beliefs and what disheartens them personally. Czarking0 (talk) 15:53, 8 October 2026 (UTC)reply
RR has been inviting opinions in the form of paper reviews (a very standard format, cf. wikt:review#Noun #2) for fifteen years now, as you could also have gleaned from the comment you are replying to. Happy to hear your arguments if you believe that this should change, but for now we'll stick with the existing format. Regards, HaeB (talk) 01:27, 9 October 2026 (UTC)reply
I think this should in part be edited down. If I ctrl+p I see 32 pages. About half of that is prose. Given that this is essentially an opinion piece I think limiting to about 5 pages of prose would be reasonable. This is especially because it is clearly AI written. AI written tends to be longer in a way that is not beneficial to the reader. Czarking0 (talk) 05:48, 18 September 2026 (UTC)reply
Consider footnote 7. Currently 93 words. It could be
Analytics data are fuzzed for privacy or unavailable for many languages. Geolocation incorrectly shows U.S., France, or Singapore for some data.
I don't think meaningful content in the 93-word footnote are lost in this proposed 20-word rewrite. Radical streamlining of more content would benefit the piece. ☆ Bri (talk) 19:23, 18 September 2026 (UTC)reply
For many reasons (data consistency and quality, external trends, and a crabload of other things)
This could be a phrase that just doesn’t exist in my dialect of English, but I’ve never heard anyone say a ‘crabload’ before, and Wiktionary gives no result for it. Is this a phrase? If not, should it be changed? Cheers, 𝔰𝔥𝔞𝔡𝔢𝔰𝔱𝔞𝔯(𝔱𝔞𝔩𝔨) (any/all) In solidarity.03:25, 24 September 2026 (UTC)reply
I just made a copyedit pass but it didn't center on concision. At this time the piece is fundamentally problematic because it is a bunch of data science[a] with some very useful observations about intervention to convert readership to editors sprinkled in, followed by what looks like a footnote about changes in governance and accelerating experiments to reverse bad trends. This isn't very readable. The data should support the central arguments, instead it looks like a research paper with some blushing personal observations. To the author: get out there more with your opinions and urgent call to action; I think this is by far the most important part of the writing. Make the research the footnotes, or links to completely different essay pages if that works better. Or as the English expression goes, don't lose sight of the forest for the trees. ☆ Bri (talk) 16:34, 24 September 2026 (UTC)reply
If I can be so bold, I've re-outlined the piece this way.
Problems
Readership is falling, masked by bot traffic
Core medium-edit-count editors are falling, masked by a temporary reprieve due time on editors' hands and lack of alternatives to online activity during COVID lockdowns
Growth only seen in high-edit-count editors (concentration risk)
Community ossification and social cost of making mistakes discourages innovation
Solutions
New governance model
New relationship to readers as a resource of future editors
Less glorification of high-edit-count contributors
New relationship to AI providers as a straight-up "truth" provider, better exposure of source qualifications
I'm not so sure about the solutions, I may have mis-understood or mis-prioritized. That's actually a problem because already at this point my time reading and re-reading is probably much higher than the average Signpost reader, who you are trying to persuade.
My advice is to lay it out very clearly as a set of really severe problems that undermine sustainability. An introduction section could lay out the problems without any research, then there could be a section that just provides a gloss of the thorough research providing evidence that those problems exist. Then get into the solutions with gusto. ☆ Bri (talk) 17:02, 24 September 2026 (UTC)reply
Notes
^Example 1: Over five pages (in my reader) are devoted to showing that about three quarters of the "desktop readers" are actually scraper bots. You could just assert that in the article, and back it up elsewhere. Example 2: Goodness of fit analysis of the models probably doesn't need to be done here.
Move these up to the appropriate position as required (e.g. adjacent to News and Notes). Copy the section header from the submission page into the |Submission= parameter so that the "Check status" button appears and works correctly.
This is from a new Signpost contributor user:1F616EMO who is an experienced Chinese language Wikipedia editor zh:user:1F616EMO~nya. I am sharing here to request early feedback. I have asked them to provide images over the placeholders in there now.
They confirmed that they wrote the piece with their own ideas but did use AI for copyediting. I do not think this needs further disclosure for AI but if anyone has another opinion then share. Bluerasberry (talk)19:22, 18 September 2026 (UTC)reply
We should pay attention to the reasonable readers, that is, those who agree to our vision and code of conduct, the five pillars, policies and guidelines, and applicable consensuses. I would suggest that the wording around this be revised, I don't think that disagreeing with some aspects of Wikipedia makes a reader unreasonable, which is heavily implied by this statement. I do think this is a good article overall, but I think the tone could be revised a bit to be less exclusionary to those with a different perspective of the encyclopedia but can nonetheless be valuable contributors to the project. Mitchsavl (talk) 01:04, 19 September 2026 (UTC)reply
Sorry that I used the wrong word. Almost all "unreasonable" should be "improper", as in WP:CONACHIEVE: "A consensus decision takes into account all of the proper concerns raised", and the implied counterpart - such a decision does not take into account any improper concerns raised. I wrote the essay with both Chinese and English Wikipedia in mind, and that resulted in awkward, non-native wordings. I will revise the essay later today to reflect this.And by writing the essay, I did not mean to exclude those who raised improper concerns as a person, but to exclude those specific concerns and, as a precaution, be extra careful to other opinions they give. Of course, they can still contribute to our movement, given that they do not insist when it is better to move on, or put something into action to prove the community wrong.The tone is intentionally exclusionary due to my Chinese Wikipedia background. On my home wiki, there is a surprisingly high tolerance to disrupters, that administrators often fail to prevent those who insist on improper concerns from further disrupting, repeat offenders are often released from a block on the sole basis of a simple promise, and the / are vote-based. The tone could be revised, but the core idea should prevail: We, as stewards of our mission, should safeguard our projects from improper demands; we may advocate and invite them into our mission, and alternatively guide them to the correct venue, but if all efforts failed, we would have no choice but to force them away. 1F616EMO (talk) 02:33, 19 September 2026 (UTC)reply
I think this should in part be edited down. If I ctrl+p I see 32 pages. About half of that is prose. Given that this is essentially an opinion piece I think limiting to about 5 pages of prose would be reasonable. This is especially because it is clearly AI written. AI written tends to be longer in a way that is not beneficial to the reader. Czarking0 (talk) 05:48, 18 September 2026 (UTC)reply
Consider footnote 7. Currently 93 words. It could be
Analytics data are fuzzed for privacy or unavailable for many languages. Geolocation incorrectly shows U.S., France, or Singapore for some data.
I don't think meaningful content in the 93-word footnote are lost in this proposed 20-word rewrite. Radical streamlining of more content would benefit the piece. ☆ Bri (talk) 19:23, 18 September 2026 (UTC)reply
For many reasons (data consistency and quality, external trends, and a crabload of other things)
This could be a phrase that just doesn’t exist in my dialect of English, but I’ve never heard anyone say a ‘crabload’ before, and Wiktionary gives no result for it. Is this a phrase? If not, should it be changed? Cheers, 𝔰𝔥𝔞𝔡𝔢𝔰𝔱𝔞𝔯(𝔱𝔞𝔩𝔨) (any/all) In solidarity.03:25, 24 September 2026 (UTC)reply
I just made a copyedit pass but it didn't center on concision. At this time the piece is fundamentally problematic because it is a bunch of data science[a] with some very useful observations about intervention to convert readership to editors sprinkled in, followed by what looks like a footnote about changes in governance and accelerating experiments to reverse bad trends. This isn't very readable. The data should support the central arguments, instead it looks like a research paper with some blushing personal observations. To the author: get out there more with your opinions and urgent call to action; I think this is by far the most important part of the writing. Make the research the footnotes, or links to completely different essay pages if that works better. Or as the English expression goes, don't lose sight of the forest for the trees. ☆ Bri (talk) 16:34, 24 September 2026 (UTC)reply
If I can be so bold, I've re-outlined the piece this way.
Problems
Readership is falling, masked by bot traffic
Core medium-edit-count editors are falling, masked by a temporary reprieve due time on editors' hands and lack of alternatives to online activity during COVID lockdowns
Growth only seen in high-edit-count editors (concentration risk)
Community ossification and social cost of making mistakes discourages innovation
Solutions
New governance model
New relationship to readers as a resource of future editors
Less glorification of high-edit-count contributors
New relationship to AI providers as a straight-up "truth" provider, better exposure of source qualifications
I'm not so sure about the solutions, I may have mis-understood or mis-prioritized. That's actually a problem because already at this point my time reading and re-reading is probably much higher than the average Signpost reader, who you are trying to persuade.
My advice is to lay it out very clearly as a set of really severe problems that undermine sustainability. An introduction section could lay out the problems without any research, then there could be a section that just provides a gloss of the thorough research providing evidence that those problems exist. Then get into the solutions with gusto. ☆ Bri (talk) 17:02, 24 September 2026 (UTC)reply
Notes
^Example 1: Over five pages (in my reader) are devoted to showing that about three quarters of the "desktop readers" are actually scraper bots. You could just assert that in the article, and back it up elsewhere. Example 2: Goodness of fit analysis of the models probably doesn't need to be done here.
