Yes, no (built-in) AI is now a feature
-
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
@libreoffice I love this boundary. If I want to use my own AI tools on my documents on my computer I want to be enabled to do that, but I will abandon any office software or browsers that insert AI into updates without a full opt-out.
-
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
@libreoffice so you're telling me that libreoffice is "Zuzai"? (https://zuzai.org)
congrats!!
-
@argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig That is what generative AI is useful for.
It is very good at generating computer code and will vastly increase the quality of what professional programmers can produce.
That is my experience
If AI is producing code of higher quality than what you can write, then that doesn't speak highly of AI; it speaks poorly of you.
As a professional programmer with 30+ years experience, I need tools that work correctly every time. AI only sometimes works correctly, and that's not good enough. I have left many tools behind because they proved unreliable, and this is one of them.
-
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
@libreoffice
🫶
Bravo -
@billyjoebowers @libreoffice I'm confused. I read their post and it implies there is no AI code in LibreOffice currently, but users can choose to install third party extensions that might interact with LLMs.
Edit: oh, I see what you mean.
-
@KittenKoder@mastodon.social That's what petrol bombs and guillotines are made for.
Tools specific to the task. -
@argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig That is what generative AI is useful for.
It is very good at generating computer code and will vastly increase the quality of what professional programmers can produce.
That is my experience
@worik @argv_minus_one @libreoffice @grvsmth @moskitokoenig
from time to time, I visit #slop fan accounts,
they all have below 3 digit followers,
that much their promotion a insane hype cycle is valued heretake the hint and consider switching platform
-
@Jeff None of the core developers use them. We rely on contributions from the wider community and don't explicitly list tools people can and can't use – but we do require people disclose when they use LLMs and generally discourage the practice. We carefully review submissions by contributors to make sure the code quality is good and that the patch author understands their own code. LLMs are prohibited for students and other mentored work.
@libreoffice hmm... where have I heard this before...
-
@grvsmth furthermore, many studies have shown that llm users perceive themselves to be more productive than they are, or even perceive productivity gains when instead their productivity is worsened
this of course doesn't acknowledge all the other consequences of llm usage to the environment, cognition, mental health, communities both local and abroad, workers rights, creators rights, global internet infrastructure, etc...
@worik @argv_minus_one @wobweger @libreoffice @moskitokoenig
-
@internic @grvsmth hm, actually my original understanding was from this study, i think:
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
but the more recent study suggest a possible improvement, though they couldn't recruit enough developers for a confident analysis:
-
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
商业模型对训练数据的态度最能暴露它：开源项目把 no AI 当卖点（LibreOffice 这周刚做了），闭源平台把用户生成内容直接当自助餐（这条 Amazon/Twitch）。一条线划在「我不放 AI 进去」，另一条划在「用户产了什么我就拿什么」。这两件事在同一周发生，挺说明问题。
-
@Jeff None of the core developers use them. We rely on contributions from the wider community and don't explicitly list tools people can and can't use – but we do require people disclose when they use LLMs and generally discourage the practice. We carefully review submissions by contributors to make sure the code quality is good and that the patch author understands their own code. LLMs are prohibited for students and other mentored work.
@libreoffice Please just ban its use.
-
@argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig That is what generative AI is useful for.
It is very good at generating computer code and will vastly increase the quality of what professional programmers can produce.
That is my experience
@worik @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
How can the: "here is the stochastic average of the training data" be of better quality?
That is not how LLMs work.The code quality is shit, especially without rigorous human supervision.
-
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
AND THAT'S WHY THEY'RE THE GOATS
-
@worik @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
How can the: "here is the stochastic average of the training data" be of better quality?
That is not how LLMs work.The code quality is shit, especially without rigorous human supervision.
@Garonenur @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
I am speaking from experience, not theory
It is not as simple as "LLMs write better code". But from now on the best code will be aided by LLMs in its construction. That is true now
-
If AI is producing code of higher quality than what you can write, then that doesn't speak highly of AI; it speaks poorly of you.
As a professional programmer with 30+ years experience, I need tools that work correctly every time. AI only sometimes works correctly, and that's not good enough. I have left many tools behind because they proved unreliable, and this is one of them.
@argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
Have you tried using a coding agent in a suitable domain? (They are not original, they are thorough)
-
@Garonenur @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
I am speaking from experience, not theory
It is not as simple as "LLMs write better code". But from now on the best code will be aided by LLMs in its construction. That is true now
@worik @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig not if we burn down every fucking fascist data center.
And no, people using LLMs are deteriorating in their skills so fast, overall code quality (which was shit anyway to begin with) will only go down. -
Yes, no (built-in) AI is now a feature
https://blog.documentfoundation.org/blog/2026/09/03/yes-no-ai-is-now-a-feature/
Thank you so much for not succumbing to the brainless "AI" cult! It means a lot to me to read principled and sensible statements like this
-
@Garonenur @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
I am speaking from experience, not theory
It is not as simple as "LLMs write better code". But from now on the best code will be aided by LLMs in its construction. That is true now
I am also speaking from experience, and my experience makes me glad that you and I haven't crossed paths professionally.
-
@worik @argv_minus_one @wobweger @libreoffice @grvsmth @moskitokoenig
How can the: "here is the stochastic average of the training data" be of better quality?
That is not how LLMs work.The code quality is shit, especially without rigorous human supervision.
Indeed. LLMs are trained on 100% of everything, and as Theodore Sturgeon once famously pointed out, 90% of everything is crap. https://en.wikipedia.org/wiki/Sturgeon%27s_law
It's a recipe for failure, and sure enough… https://www.pcmag.com/news/vibe-coding-fiasco-replite-ai-agent-goes-rogue-deletes-company-database
