Board of Directors Meeting 2026-10-26 - Call Minutes

Board of Directors Meeting 2026-10-26 - Call Minutes

Date: Monday, October 26, 2026
Location: Jitsi

Session chair: Eliane Domingos de Sousa
Notes for the minutes: Florian Effenberger, Mike Saunders

In the call:

Board - Eliane, Mike, Sophie, Paolo, Laszlo, Osvaldo
Board deputies - none
Membership committee - Andreas
Membership committee substitutes -
Team - Florian, Ilmari, Jonathan, Aszlig
Community - Michael Meeks

Simon sent apologies for absence before and could not attend the meeting.

Representation: none

The Board of Directors at the time of the call consists of 7 seat holders without deputies. In order to be quorate, the call needs to have 1/2 of the Board of Directors members, which gives 4. A total of 6 Board of Directors members are attending the call.

The board waives all formal statutory requirements, or requirements in the foundations articles, or other requirements regarding form and invitation, time limits, and for the topics discussed in this meeting.

The call is quorate. From now on, motions can be passed with the agreement of a simple majority of those remaining present. The majority threshold is currently 4.

The meeting commenced at 17:05 UTC / 18:05 CET, Berlin time

A. Public section

  • Welcome to our new team member Aszlig (system administrator)

    • Hi! Usually involved in FOSS, NixOS, distribution stuff, have private projects doing infra etc. (Aszlig)
    • Thanks to Guilhem and Cloph for onboarding (Mike)
    • Already got into testing the infra and studying it (Aszlig)
  • AI and LibreOffice

    • Laszlo writes in chat: “I suggest to talk about the future, especially in light of this Register article, and the referred open source software, an AI reverse engineered AI clone of Microsoft Word (and all the Adobe products): Rust upstart reveals 'reimplementation' of Microsoft Office – with Claude on coding duty - Given some checking (the code base and the repository of “WordCraft”), and some testing (only the frighteningly function-rich and fast, GPU accelerated Adobe InDesign clone “DesignCraft”, yet – on Linux!), also seeing the pace of the open source AI development, I find the situation quite concerning. These are the kinds of software where users can add what they’re missing in minutes, rather than over the course of years (10–20 years or never). Combined with the April 1st event, i.e. losing our minimally necessary developer base, it seems for me, it’s like a low blow for TDF and the future of LibreOffice. If not the problem is the acceleration pof the AI devs found in LibreOffice or Microsoft Office. Even if they don’t function with 100% perfection (but already there is a paragraph composer in DesignCraft – not in Scribus yet), for a significant (perhaps even overwhelming?) portion of our micro-donors, they will create a better office suite than LibreOffice for ensuring compatibility with MSO. And we are not just talking about cloning, but about everything else they will be capable of – things our users have only dreamed of until now…”
    • Understand the context (Jonathan)
      • recently a few independent devs using commercial LLMs to generate large apps
      • one dev made a clone of Adobe products, and cloning MS Office
      • became a viral hit on the internet in recent weeks
      • interesting and exciting that LLMs are capable of performing at this level
      • but looking at the source code - it’s low quality. Professionals confirm. The apps look the same, but are extremely incomplete and unreliable. Not engineered in a way that’s sustainable, even if you wanted to invest in moving them forward
      • intellectually fascinating, but need to manage expectations
      • we’ve received a lot of contributions from potential contributors who don’t have a strong background. Patches that they submit are consuming a lot of time - very large and complicated, superficially solve issues, but aren’t obvious
    • DesignCraft - I looked at code in detail (Aszlig)
      • wouldn’t call it a proper product. would work for a few people, but a lot of code that I’d usually call debt
    • LLVM project has their AI policy where they talk about “extractive contributions” (Ilmari)
      • they take a lot of time from contributors
      • talking about updating our AI policy. have to ask for license statement, tell people to get into the normal contribution flow
      • trying to build a community - seems strange to build a community of machines. doesn’t make sense in the context of a foundation
      • add item to policy: people have to submit a certain number of patches without AI, so after that we have confidence to accept LLM-assisted patches from them
    • we have the question of the educational purpose (Paolo)
      • no learning process if people submit AI patches
      • issue is already hitting the codebase
      • question is: are there areas where “stochastic parrots”, LLMs, where we could get something useful out of it? bugs, processes we can speed up?
    • I’m planning on spending Dec vacation experimenting with agentic development, hoping to come back with some best practices (Jonathan)
    • which area? Still back to using the same tool (Paolo) - what about the purpose of the project - where do we put the threshold
    • identified a number of representative tasks for my experiment, to see how it works or fails (Jonathan)
      • still discussing in the team about our mentoring
    • what works quite well is code reviews (Aszlig)
      • some false positives, but works pretty well, at least in my projects
    • have been receiving lots of vulnerability reports, and AI companies may have been involved. Can’t deny we are getting benefits in that way (Ilmari)
  • Community Q&A: 10 minutes

    • hoping to get a response shortly (Michael)
      • ball is in your court - want to check we’re not blocking on our side
    • will have a private meeting about this today (Eliane)

B. Trustee/members-only section

[REDACTED: shared with members]

C. Board and MC only section

[Andreas Mantke joined at 17:32 UTC / 18:32 CET]

  • board member tax data needed for main account [see mail from Stephan 17:02 Berlin time, 26 October]

  • [REDACTED: 36 lines of legal discussion]

The meeting ended at 17:55 UTC / 18:55 CET, Berlin time.

Eliane Domingos de Sousa (Session chair)
Mike Saunders (Keeper of the minutes)

1 Like

Hi,
I think this discussion also highlights challenges that existed long before AI: the considerable amount of implicit knowledge behind almost every commit, the limited availability of experienced developers for mentoring, the learning syllabus that still needs to be built, and language barriers.

We should also remember that not every potential contributor is a professional developer, nor does everyone have twenty years ahead of them to become one. This should not mean they have nothing valuable to contribute.

Perhaps the real opportunity for AI is not simply to generate more code, but to make knowledge more accessible, lower the barriers to contribution, and help build a more inclusive and sustainable community.

even the non-transparency gets more and more obfuscated :frowning: :face_with_thermometer:

definitely ununderstandable why these 33 lines are not public ?! :thinking:

The reason is simple: It’s the first time in quite a while that we have a members-only section in the meeting. We didn’t discuss so far whether these contents can be public. From my side, absolutely - but as members speak out and are minuted, they might not like to have it public without being aware of it before. To not expose anyone without their consent, I did not publish this part for now.

However, I agree with you insofar that we should discuss at the next meeting or on the tdf-internal list if members are in general fine with having the members-only section published. I’ll send an e-mail right now.

expelled or not, seems these minutes are is absolutely no way an effort of any transparency to the community

@mmeeks, feel free to remind the community what response is that …

@elianedomingos, private to whom ?

to the happy few who can understand, no doubt.
for the community, be sure it’s just plain opaque, arbitrary, random.
thus it’s non existence.

huh?
so there’s more than the one sent @tdf-internal 1:01 PM ? :thinking:

I honestly begin to get upset by the way you diminish any work that is being done. You are permanently criticizing in non nice words what others do. How would you feel if others did the same to you? It is simply not nice and it is simply not appropriate, and I ask you to reconsider your behaviour in this very forum, so we can have a constructive discussion. If you are not interested in a constructive discussion, please let me know.

No. The part was published to tdf-members in full (not tdf-internal as you say). I do not want to publish it to the general public unless members are fine we do so.

There was a private discussion about a legal matter in the board-only part of the call.

So, you now not only diminish the work of others, but even claim it doesn’t exist. That’s a new level. I wonder if you have any interest in a constructive discussion, or just want to be vocal. If the latter, this is the wrong community.

Essentially, for the same reason that the BoD-only section is redacted without a specific embargo decision with a disclosure date: Because even though that’s supposed to be impossible - it’s convenient to redact things, and incovenient to have written record of things. And since we don’t take our BoD to task about such things, illict secrecy continues as common practice.

1 Like

While for sure we can always improve, please let’s not forget we published a lot of previously private minutes. Legal parts were redacted, but a lot that was kept private in the past has been made public - so it’s not true that everything stays private eternally. Again, we can always improve, but it should be visible TDF is working on this.

Bugzilla has 88 tickets tagged by me as easyhack with skillDesign that come with code pointer and explanation about how to fix the issue. There are also 7 easyhacks with skillCpp and difficultyBeginner.

You do not need to a degree in IT to contribute, and I disagree with the notion of “limited availability of experienced developers” - while this always is true, TDF has people who care only about mentoring. What other open-source project does the same?

Sure, I understand the wish for more experience developers and more… whatever. And do ask for that too. But when we estimate the success of the project and how well we prosper we can easily watch from another perspective. And my take on the PITA AI is that it rather kills our fine project.

If me was the (hopefully benevolent) dictator in design (with sufficient power to realize ideas), the look and feel of the application will become more clean and more pleasing. But the chance for everyone to be involved, to share different visions, to talk about strategies makes open source and LibreOffice a lighthouse project. Enslaving yourself to AI would be the same mistake - development becomes faster, sloppier, and. you lose expertise. AI is unethical and environmentally a catastrophe.

I’m afraid we talk about the same things with different point of view.

Ideally, we should help contributors progress through several stages:

  1. Guidance: I know where to make a change in the code.
  2. Understanding: I understand why this change works.
  3. Broader knowledge: I understand the components involved and how they interact.
  4. Autonomy: I can independently identify where changes need to be made.
  5. Expertise: I can design and implement new features without constantly relying on experienced developers.

Easyhacks are valuable for the first stage, but I believe we also need to consider how we support contributors throughout the following stages.
As far as I can see, mentoring activities are rarely documented or made publicly available in a way that would benefit future contributors. Much of the knowledge shared during these interactions remains inaccessible to others.

This is one of the reasons why people turn to AI. Despite its limitations and the need to carefully verify its answers, AI can provide enough guidance to help contributors navigate unfamiliar parts of the codebase and make progress independently.

The UI configuration files are a good example. Their underlying mechanisms are quite specific, yet there is very little accessible documentation explaining why they work the way they do. (which link between uic , glade ,UNO IDL, etc. ?)

I don’t see AI as a replacement for mentoring, but rather as a way to compensate, at least partially, for the lack of accessible and structured knowledge.

If you change “afraid” into “glad to” (have different POVs) and “should help contributors” into “improve helping contributors” we are on the same page. Me and likely most others would sign your 5 points and reject AI for this reason. You are welcome to suggest how to propose a solution like “no AI on easyhacks”. But I’m afraid we are hi-jacking the BoD meeting minutes :-).

Are you aware of this new article, which will hopefully be expanded also from the perspective of different platforms: Development/Navigating the LibreOffice codebase - The Document Foundation Wiki ?

concretely : https://ask.libreoffice.org/t/deepwiki-devin-ai-as-assistant-in-my-questions-open-points-around-use-problems-with-writer/131977