Talk:Abstract Wikipedia

Archives

---

1, 2, 3, 4

Architect, a working solution, rejected by some representant of WMF: The wrap (going away)

My answer to @Sannita (WMF) , offered here for the community, for transparency.

Sannita answer to me seeking collaboration to integrate Architect:

"HI Réjean,

first of all, sorry for the late reply. We've been very busy with the launch of Abstract Wikipedia as a project, but we found time to take a look at your project.

Unfortunately, we're not sure your solution would work with what we have in mind. We are happy that you decided to share the project with us and with the wider community, though, and we wish you all the success you deserve for your project.

Cheers,

L.

"

My answer:

Thanks for the candid note.

I understand your decision, and I’m not going to try to litigate taste. If Architect isn’t what you had in mind, then the answer is simply “no,” and that’s fine.

That said, I want to be precise about what you’re declining, because “not what we had in mind” can mean anything from “we don’t want this capability” to “we don’t want to own the integration work.” Architect is not a concept deck. It’s a working system that already does the hard, measurable part of the Abstract Wikipedia promise: take structured meaning and reliably realize it into grammatical sentences across many languages, fast, deterministically, and with an audit trail (AST/debug trace) that supports regression testing and controlled iteration. That’s not a speculative direction; it’s functioning end-to-end.

If what you were expecting was a fully MediaWiki-native, community-governed, revision-history-integrated authoring platform that lives entirely inside the Wikimedia ecosystem, then yes—Architect is not that. It is a renderer/workbench with its own UI and toolchain, because it was built to prove capability and iterate rapidly on coverage (lexicon, grammar, QA) rather than to reimplement the full Wikimedia process stack on day one. Those are different scopes. Conflating them is a category error: an integration layer is not the core engine, and insisting the engine be born inside one specific host environment is a constraint choice, not a definition of the problem.

I didn’t “overcomplicate” the system. I did the opposite: I reduced the problem to the thing that must be true before anything else matters—does abstract content actually render into many languages in a controllable way, and can you measure/extend coverage without hand-waving? Architect answers that with running code, reproducible outputs, and tooling that makes gaps visible and fixable.

If your refusal is ultimately about ownership and governance—who hosts it, who maintains it, how community review works, what canonical storage format is, and what is acceptable operationally—those are valid institutional questions. They’re also separate from whether the renderer works. Architect works. The remaining questions are product/process decisions that any implementation must eventually face.

I’m going to keep using Architect and continuing development regardless, because the capability is useful with or without Wikimedia. In fact, given the mismatch in expectations, I’m planning my next steps outside the Wikimedia ecosystem, where the constraints are clearer and the integration surface can be defined by the needs of the users actually shipping content.

If anyone on your side is still curious, I’ll publish a short demonstration of what works (multilingual realization from structured meaning, AST auditability, and coverage/QA tooling) as part of that migration. You’re welcome to watch it when it’s ready—not as a pitch, not as a request for approval, just as a factual record of what exists.

No action required on your end.

I will soon update https://github.com/Rejean-McCormick/abstract-wiki-architect/wiki

I must confess I was hoping to find a likeminded community here, willing to break language barrier in order to build a more inclusive world. Well, breaking language barrier is the goal, so I guess I'm confused, or some WMF foundation representatives are confused. Still, I was here to answer questions about Architect, here to collaborate, but I mostly hit a wall, maybe because I don't express myself like most do. Réjean McCormick (talk) 21:15, 23 February 2026 (UTC)

Hi @Sannita (WMF), although he said "no action required on your end", per WP:ADMINACCT, I would like to receive an adequate reason, so the community can respond. I can't do WP:CRYSTAL, perhaps WMF prefers community translation instead of machine translation, or finds the proposal technically deficient, I have no idea, I am no expert, but I'd like the official assessment, given this appears to be major proposal, thank you. ~2026-13728-50 (talk) 19:27, 3 March 2026 (UTC)

We decided to go another way, to put it simply. This is to be treated as a community proposal that just didn't get approval. No action required on anybody's end. Sannita (WMF) (talk) 09:47, 4 March 2026 (UTC)

I have to say, that response feels incredibly dismissive.

I was not merely suggesting a theoretical path forward; I built a working solution. Asking the community to duplicate this effort from scratch—without providing any technical justification—is counterproductive.

There is currently no complete roadmap for Abstract Wiki. I brought forward a viable solution that hadn't been considered, and it feels it was rejected simply because the team wasn't involved in the underlying technological choices. I reached out to collaborate, so the opportunity was always there.

The current feedback essentially boils down to: the project doesn't yet know how to get Abstract Wiki fully working, but is rejecting a functioning solution because it doesn't align with a plan that doesn't exist yet.

I made a point to integrate the existing parts of your solution. I didn't come to overthrow your work; I just avoided building on an unworkable foundation. I relied extensively on AI to analyze and select the best technical architecture, and while I extended multiple invitations for you to join that process, you did not participate.

~~~~ Réjean McCormick (talk) 11:51, 4 March 2026 (UTC)

For those wondering about the "behind the scenes," I know I arrived overly enthusiastic, but what I built fully justifies it. I am not a PR person or a salesperson. I don't believe I should be dismissed for having a little fun when my technological solution is incredibly sound.

I feel you don't want to give me credit for what I've built simply because I'm not part of your community. It seems it would put you to shame to admit I accomplished what you couldn't. But this isn't an ego war—or at least, it shouldn't be.

I avoid duplicating work, and I don't compete. Yet, the leadership here is asking you all to do exactly that: maintaining the effort to build an abstract wiki when I am already miles ahead.

I kept my extracurricular activities entirely separate from this initiative, so ultimately, you have no valid reason to reject what I delivered.

A lot of hard work went into this, and I am deeply disappointed in the leadership here. Réjean McCormick (talk) 12:08, 4 March 2026 (UTC)

I also think this response is unsatisfying and insufficient as well as somewhat disrespectful of the user's time and effort. One potential good reason for not specifying things however could be if doing so may be a problem to adoption later on where it would shape the community view of this but I doubt that's the reason or that specifying things a bit more would be a problem.

The tool and idea sounded like it could be very useful. It was difficult to make much sense of it, sounded like a lot of buzzwords, and most importantly had no concrete demo by which one can understand it and see its usefulness.

I hope the user is not demotivated by this and considers picking it up again when Abstract Wikipedia is at a more advanced stage, particularly by implementing a demo (online tool or video) that illustrates the usefulness. Maybe AW is not even viable – in terms of ever becoming truly useful and used a lot – without a tool like the one proposed here. The idea of the tool while I may have misunderstood it challenged my view that Abstract Wikipedia will never be more than useless super-short natural language sentences of Wikidata statements but how it went furthers my assessment that Abstract Wikipedia is rather unlikely to ever become a promising new project that gets used a lot and encourages further work on the Machine Translation Project that I proposed which is an entirely different approach to what AW in terms of real-world-use seems to be about. Prototyperspective (talk) 12:08, 4 March 2026 (UTC)

@Prototyperspective I understand how specialized language can seem like buzzwords. But you can give my doc to your favorite AI—these are definitely not buzzwords.

Regarding the demo, I would have appreciated it if leadership had asked me about my current progress instead of giving an empty answer, closing the door for collaboration. Réjean McCormick (talk) 12:18, 4 March 2026 (UTC)

Concern

One concern I have about this is that it appears to rely on the concept of a 'universal article', but assuming each entry will be based on the highest quality version, most will be copies of en.wiki articles. The point of having a wikipedia in your own language is that the content is written by people from your culture for people from your culture, prioritising native-language sources (ie. perspectives). The example entry given, a year whose only event is a French invasion of Corsica, is emblematic of this problem. This looks like it's going to be extremely Eurocentric/Anglocentric. The only solution I can see is have abstract entries be bare-bone (stubs even) with no cultural persuasion, but even that doesn't seem feasible.

I'm not going to pretend I understand this at a technical level, but another concern is that words/phrases in different languages rarely have exact matches, with very high conceptual diversity among the worlds' languages/cultures. There is simply no way to have a 'universal language' that people work in, nor is there a way to communicate/translate concepts and ideas without using some sort of language/medium.

Even if my above concerns are unfounded, I'm very worried this is inevitably going to go down like a sack of potatoes when rolled out/presented and become another 'scandal'. Please get communities on board before putting more resources into this. Kowal2701 (talk) 23:01, 4 March 2026 (UTC)

@Kowal2701 Thank you for airing your concerns. We will share them with the community of the project, of course, and we will stay vigilant on this. Our main objective is to share more knowledge in more languages, and of course this will happen with a combination of efforts from more communities, not just the English one. We are also aware that there is a problem of translation of words between languages, but it is a relative problem, since the abstract articles will be a start for people to work with, and they are not supposed to substitute in entirety (and definitely not forever) a human-made article. But we are very aware of the problems you raise, and we're working to avoid or reduce such impact as much as possible. Sannita (WMF) (talk) 09:42, 5 March 2026 (UTC)

One approach might be identifying topics that are unique or near-unique to each language edition, and ensuring that Abstract Wikipedia has a cross-section of those early on. Some of those won't be relevant to other language editions, but other may identify areas that have insufficient coverage in the largest editions. Arlo Barnes (talk) 23:06, 5 March 2026 (UTC)

That sounds good, though it’s also how en.wiki's lead for en:house focuses on Western conceptions and North America, that sort of thing that largely arises from a) English-language sources having that focus, and b) most en.wiki editors being Western/American Kowal2701 (talk) 23:15, 5 March 2026 (UTC)