AI generated mails

youkidearitai

9 days ago

2026年8月23日(日) 22:24 Christian Schneider <cschneid@cschneid.com>:

>
> In the thead "[PHP-DEV] [RFC] [Discussion] array_match" I got a reply starting with "Thanks for your feedback and for raising these valid points" and then sort of reiterating things I already addressed ("requires adding delimiters and wrapping the needle in `preg_quote()`") which leads me to believe that the response was AI generated. I might be wrong in this specific case but it got me thinking.
>
> I'm not sure I want to spend time arguing with AIs and was considering replying with "AI;DR" (AI; didn't reply) but I realize that this can be considered rude.
>
> Just to be clear: I realize that AIs help a lot for non-native speakers to express themselves but when I get the feeling that the whole argument was generated and not only translated then I get a bit defensive about my personal time and energy.
>
> As I'm expecting more and more generated mails on this list: What is the feeling on the mailing list about how to handle this?
>
> Regards,
> - Chris

Hi, Chris

From a Japanese not native English speaker,
I using DeepL(or Google translator) when since they released.
So I don't think reject unconditionally.

(One another note, He is maybe from RtoL language speaker, So I would
like understand his know-how)

However, That matter was a little bit angry when he not reading the link,
so I can understand your emotions too.

Anyway, I know that he using the AI, but I would like get his
language's know-how.

If you create the rule, So I'll leave it to you.

Regards
Yuya

--
---------------------------
Yuya Hamada (tekimen)
- https://tekitoh-memdhoi.info
- https://github.com/youkidearitai
-----------------------------

سپهر محمودی

9 days ago

در تاریخ یکشنبه ۲۳ اوت ۲۰۲۶، ۱۶:۵۲ Christian Schneider <
cschneid@cschneid.com> نوشت:

> In the thead "[PHP-DEV] [RFC] [Discussion] array_match" I got a reply
> starting with "Thanks for your feedback and for raising these valid points"
> and then sort of reiterating things I already addressed ("requires adding
> delimiters and wrapping the needle in `preg_quote()`") which leads me to
> believe that the response was AI generated. I might be wrong in this
> specific case but it got me thinking.
>
> I'm not sure I want to spend time arguing with AIs and was considering
> replying with "AI;DR" (AI; didn't reply) but I realize that this can be
> considered rude.
>
> Just to be clear: I realize that AIs help a lot for non-native speakers to
> express themselves but when I get the feeling that the whole argument was
> generated and not only translated then I get a bit defensive about my
> personal time and energy.
>
> As I'm expecting more and more generated mails on this list: What is the
> feeling on the mailing list about how to handle this?
>
> Regards,
> - Chris
>

---------------

Hi everyone,

Chris, just to clear things up: I am a real human being and I know my way
around PHP.

If my previous emails sounded overly formal, it was purely out of deep
respect for the experts on this list. Otherwise, I can easily communicate
in a much more direct and straightforward tone, just like I'm doing right
now.

Also, regarding the mixed-up quotes and repetitions you mentioned—do you
really think an AI would make those specific kinds of messy mistakes? If
anything, that actually proves I'm a real person!

Going forward, I'll drop the ultra-formal language and just use this normal
tone with you all.

Best regards,
Sepehr

Christian Schneider

9 days ago

Am 23.08.2026 um 17:13 schrieb ⁨سپهر محمودی⁩ <⁨sepehrphpr@gmail.com⁩>:

> Chris, just to clear things up: I am a real human being and I know my way around PHP.

I didn't doubt that you are human but I want to encourage people to keep mails succinct and to the point as mailing lists act as multipliers: One person writes a message for hundreds or thousands to spend time reading.

We've seen GitHub being overwhelmed by the sheer amount of PRs during the last months, and AI making it easier to generate those is IMHO a big part of the reason for this. I Want to avoid internals being overwhelmed by long back and forth threads due to the ease of writing replies using AI.

Apologies if you took my email personal, this thread was not meant to be about any specific email but about a technology and how we handle it to keep internals working.

> If my previous emails sounded overly formal, it was purely out of deep respect for the experts on this list. Otherwise, I can easily communicate in a much more direct and straightforward tone, just like I'm doing right now.

Yes, please use direct language, no extra politeness needed for technical discussions like this, certainly not for me ;-)
Additionally it saves time for readers if you strip down the quoted text to the relevant parts.

Regards,
- Chris

سپهر محمودی

8 days ago

در تاریخ یکشنبه ۲۳ اوت ۲۰۲۶، ۱۹:۱۷ Christian Schneider <
cschneid@cschneid.com> نوشت:

> Am 23.08.2026 um 17:13 schrieb ⁨سپهر محمودی⁩ <⁨sepehrphpr@gmail.com⁩>:
> > Chris, just to clear things up: I am a real human being and I know my
> way around PHP.
>
> I didn't doubt that you are human but I want to encourage people to keep
> mails succinct and to the point as mailing lists act as multipliers: One
> person writes a message for hundreds or thousands to spend time reading.
>
> We've seen GitHub being overwhelmed by the sheer amount of PRs during the
> last months, and AI making it easier to generate those is IMHO a big part
> of the reason for this. I Want to avoid internals being overwhelmed by long
> back and forth threads due to the ease of writing replies using AI.
>
> Apologies if you took my email personal, this thread was not meant to be
> about any specific email but about a technology and how we handle it to
> keep internals working.
>
> > If my previous emails sounded overly formal, it was purely out of deep
> respect for the experts on this list. Otherwise, I can easily communicate
> in a much more direct and straightforward tone, just like I'm doing right
> now.
>
> Yes, please use direct language, no extra politeness needed for technical
> discussions like this, certainly not for me ;-)
> Additionally it saves time for readers if you strip down the quoted text
> to the relevant parts.
>
> Regards,
> - Chris
>

-----

Hi Chris,

Thanks for the clarification. No hard feelings at all! I totally agree with
keeping things direct and focusing on the technical side.

Also, thanks for the tip about trimming the quoted text—I’ll definitely
make sure to do that from now on.

Cheers,
Sepehr

Osama Aldemeery

9 days ago

On Sun, Aug 23, 2026 at 4:21 PM Christian Schneider <cschneid@cschneid.com>
wrote:

>
> In the thead "[PHP-DEV] [RFC] [Discussion] array_match" I got a reply

starting with "Thanks for your feedback and for raising these valid points"
and then sort of reiterating things I already addressed ("requires adding
delimiters and wrapping the needle in `preg_quote()`") which leads me to
believe that the response was AI generated. I might be wrong in this
specific case but it got me thinking.

>
> I'm not sure I want to spend time arguing with AIs and was considering

replying with "AI;DR" (AI; didn't reply) but I realize that this can be
considered rude.

>
> Just to be clear: I realize that AIs help a lot for non-native speakers

to express themselves but when I get the feeling that the whole argument
was generated and not only translated then I get a bit defensive about my
personal time and energy.

>
> As I'm expecting more and more generated mails on this list: What is the

feeling on the mailing list about how to handle this?

>
> Regards,
> - Chris

Hi Chris,
Thank you for bringing this up.

Since this isn't the first time this has come up, I'd like to ask whether
it's time internals discussed an AI-usage policy?

A written document would tell contributors what is and isn't acceptable
here. Whether that is a disccussion, an RFC, or a PR.
And it would save everyeone from settling it case by case each time.

It is not new ground either...Python, CURL, Linux Kernel, and Fedora (among
others) have all faced the same issue, and each landed on a policy.

For reference:
Python: https://devguide.python.org/getting-started/generative-ai
curl: https://curl.se/dev/contribute.html#on-ai-use-in-curl
Linux kernel: https://docs.kernel.org/process/coding-assistants.html
Fedora:
https://docs.fedoraproject.org/en-US/council/policy/ai-contribution-policy

Thanks,
Osama

Christian Schneider

9 days ago

Am 23.08.2026 um 20:41 schrieb Osama Aldemeery <aldemeery@gmail.com>:

> Since this isn't the first time this has come up, I'd like to ask whether it's time internals discussed an AI-usage policy?
>
> A written document would tell contributors what is and isn't acceptable here. Whether that is a disccussion, an RFC, or a PR.
> And it would save everyeone from settling it case by case each time.
>
> It is not new ground either...Python, CURL, Linux Kernel, and Fedora (among others) have all faced the same issue, and each landed on a policy.
>
> For reference:
> Python: https://devguide.python.org/getting-started/generative-ai
> curl: https://curl.se/dev/contribute.html#on-ai-use-in-curl
> Linux kernel: https://docs.kernel.org/process/coding-assistants.html
> Fedora: https://docs.fedoraproject.org/en-US/council/policy/ai-contribution-policy

One little note here: The above seems to concern code contributions like PRs.
I was more talking about the discussions here on the internals mailing list.

But in the end it applies to both: How do we handle the higher amount of generated code and mails.
And how can guidelines hopefully somewhat protect the limit resources of the PHP maintainers?

Regards,
- Chris

Osama Aldemeery

8 days ago

On Sun, Aug 23, 2026 at 9:10 PM Christian Schneider <cschneid@cschneid.com>
wrote:

>
> Am 23.08.2026 um 20:41 schrieb Osama Aldemeery <aldemeery@gmail.com>:
> > Since this isn't the first time this has come up, I'd like to ask

whether it's time internals discussed an AI-usage policy?

> >
> > A written document would tell contributors what is and isn't acceptable

here. Whether that is a disccussion, an RFC, or a PR.

> > And it would save everyeone from settling it case by case each time.
> >
> > It is not new ground either...Python, CURL, Linux Kernel, and Fedora

(among others) have all faced the same issue, and each landed on a policy.

> >
> > For reference:
> > Python: https://devguide.python.org/getting-started/generative-ai
> > curl: https://curl.se/dev/contribute.html#on-ai-use-in-curl
> > Linux kernel: https://docs.kernel.org/process/coding-assistants.html
> > Fedora:

https://docs.fedoraproject.org/en-US/council/policy/ai-contribution-policy

>
>
> One little note here: The above seems to concern code contributions like

PRs.

> I was more talking about the discussions here on the internals mailing

list.

>
> But in the end it applies to both: How do we handle the higher amount of

generated code and mails.

> And how can guidelines hopefully somewhat protect the limit resources of

the PHP maintainers?

>
> Regards,
> - Chris

Hi Chris,

Yes, and Yes.

The examples I gave were really just to make the point that others have
landed policies, not to claim they map neatly onto the discussion side.

And your two questions are exactly what I was getting at.

With AI probably not going anywhere in the foreseeable future, I think it
would be better for everyone to have rules that define what is and isn't
acceptable in dicussions, RFCs, and PRs.
What I was suggesting is a written policy on AI usage...the same route
others have already taken.

Now I recognize it might be a bit odd for a non-maintainer like me to raise
a process/governance question...but I have seen people bring up the same
issue
in different clothes over the months. So yeah...I think it's worth writing
something down.

Regards,
Osama

Pierre Joye

7 days ago

hi again:)


On Mon, Aug 24, 2026, 1:49 AM Christian Schneider <cschneid@cschneid.com>
wrote:

> Am 23.08.2026 um 20:41 schrieb Osama Aldemeery <aldemeery@gmail.com>:
> > Since this isn't the first time this has come up, I'd like to ask
> whether it's time internals discussed an AI-usage policy?
> >
> > A written document would tell contributors what is and isn't acceptable
> here. Whether that is a disccussion, an RFC, or a PR.
> > And it would save everyeone from settling it case by case each time.
> >
> > It is not new ground either...Python, CURL, Linux Kernel, and Fedora
> (among others) have all faced the same issue, and each landed on a policy.
> >
> > For reference:
> > Python: https://devguide.python.org/getting-started/generative-ai
> > curl: https://curl.se/dev/contribute.html#on-ai-use-in-curl
> > Linux kernel: https://docs.kernel.org/process/coding-assistants.html
> > Fedora:
> https://docs.fedoraproject.org/en-US/council/policy/ai-contribution-policy
>
>
> One little note here: The above seems to concern code contributions like
> PRs.
> I was more talking about the discussions here on the internals mailing
> list.
>
> But in the end it applies to both: How do we handle the higher amount of
> generated code and mails.
> And how can guidelines hopefully somewhat protect the limit resources of
> the PHP maintainers?
>

Most if not all security i could see are discovered by AI. I can imagine
many PRs are, at least, partially AI generated too.

One issue I mentioned was the process (I need another word as this not a
process anymore at this stage), and will, or ability, to delegate reviews
of PRs.

We cannot expect RMs, or RFC policies checkers to have the knowledge for
everything in php-src. The latter, and delegation, is harder to solve than
the policies.

I only we don't even think about an anti AI policy or similar bad ideas :)

cheers,

--
Pierre

@pierrejoye

Derick Rethans

7 days ago

On Sun, 23 Aug 2026, Osama Aldemeery wrote:

> On Sun, Aug 23, 2026 at 4:21 PM Christian Schneider <cschneid@cschneid.com>
> wrote:
> >
> > In the thead "[PHP-DEV] [RFC] [Discussion] array_match" I got a
> > reply starting with "Thanks for your feedback and for raising these
> > valid points" and then sort of reiterating things I already
> > addressed ("requires adding delimiters and wrapping the needle in
> > `preg_quote()`") which leads me to believe that the response was AI
> > generated. I might be wrong in this specific case but it got me
> > thinking.
> >
> > I'm not sure I want to spend time arguing with AIs and was
> > considering replying with "AI;DR" (AI; didn't reply) but I realize
> > that this can be considered rude.

I think that replying with "AI;DR" is less rude than expecting all 1199
receivers on this list to want to read overly long-winded LLM generated
text, where seamingly the human-written text from one contributor gets
fed into a sloperator to get a reply.

> > Just to be clear: I realize that AIs help a lot for non-native
> > speakers to express themselves but when I get the feeling that the
> > whole argument was generated and not only translated then I get a
> > bit defensive about my personal time and energy.
> >
> > As I'm expecting more and more generated mails on this list: What is
> > the feeling on the mailing list about how to handle this?
>
> Since this isn't the first time this has come up, I'd like to ask
> whether it's time internals discussed an AI-usage policy?

Yes, this is long overdue.

For now, I will talk about this mailinglist first (and only).

With that, we end up with drivel like this:
https://externals.io/message/132300#132332 (au contraire the acceptance
to not do this, as per https://externals.io/message/132300#132317).

This is a list for *people* to discuss the PHP language, its features,
and its policies.

Therefore, a policy for this list ought to be that you *have* to write
the content yourself. If that means writing it in non-English and
feeding it through a *translating-LLM* only, that that seems reasonable.

What does not seem reasonable is to ask a tool to write the text for
you, and then copy and pasting this as a reply to this list — especially
because it doesn't deal with the trimming and quoting (as per our
guidelines at
https://github.com/php/php-src/blob/master/docs/mailinglist-rules.md ).

cheers,
Derick

Larry Garfield

7 days ago

On Tue, Aug 25, 2026, at 10:25 AM, Derick Rethans wrote:

>> > As I'm expecting more and more generated mails on this list: What is
>> > the feeling on the mailing list about how to handle this?
>>
>> Since this isn't the first time this has come up, I'd like to ask
>> whether it's time internals discussed an AI-usage policy?
>
> Yes, this is long overdue.
>
> For now, I will talk about this mailinglist first (and only).
>
> With that, we end up with drivel like this:
> https://externals.io/message/132300#132332 (au contraire the acceptance
> to not do this, as per https://externals.io/message/132300#132317).
>
> This is a list for *people* to discuss the PHP language, its features,
> and its policies.
>
> Therefore, a policy for this list ought to be that you *have* to write
> the content yourself. If that means writing it in non-English and
> feeding it through a *translating-LLM* only, that that seems reasonable.
>
> What does not seem reasonable is to ask a tool to write the text for
> you, and then copy and pasting this as a reply to this list — especially
> because it doesn't deal with the trimming and quoting (as per our
> guidelines at
> https://github.com/php/php-src/blob/master/docs/mailinglist-rules.md ).
>
> cheers,
> Derick

I agree with Derick.  Using a translator to assist with a language barrier is reasonable, and these days they've all been replaced with LLMs (justified or not).  But I would still prefer to have that explicitly stated, so we know what we're dealing with.  (And can use simpler English in replies, potentially, to aid with understanding.)

Beyond that, I want to talk to humans, not slop machines.  If someone's diplomatic skills are not up to the task of discussing on the list without filtering through a chatbot, then they need to work on their diplomatic skills before subjecting us to slop.

--Larry Garfield

Pierre Joye

7 days ago

hello,

On Tue, Aug 25, 2026, 10:36 PM Larry Garfield <larry@garfieldtech.com>
wrote:

>
>
> I agree with Derick.  Using a translator to assist with a language barrier
> is reasonable, and these days they've all been replaced with LLMs
> (justified or not).  But I would still prefer to have that explicitly
> stated, so we know what we're dealing with.  (And can use simpler English
> in replies, potentially, to aid with understanding.)
>
> Beyond that, I want to talk to humans, not slop machines.  If someone's
> diplomatic skills are not up to the task of discussing on the list without
> filtering through a chatbot, then they need to work on their diplomatic
> skills before subjecting us to slop.
>

I have to interfere here again.

I explained a perfectly valid case where llm are used, and they are many
orders of magnitude better than translators, even based on llm.

I experience that daily. And the primary reason the personsI work with do
it is, guess it, to respect me or others.

To say here, given someone has been directly pointed to via his post, would
be extremely bad taken in many cultures (read: insulting, losing face, and
they will move away).

As the "official" language is English here and most oss project, I cannot
let that go unnoticed.

We must be welcoming for non natice speakers. And for one, I welcome them.
And if the price is to have a slightly longer mail at time,  so be it.
respect and welcoming must prevail.

--
Pierre

@pierrejoye

Osama Aldemeery

6 days ago

On Tue, Aug 25, 2026 at 7:30 PM Pierre Joye <pierre.php@gmail.com> wrote:

>
> hello,
>
>
>
> On Tue, Aug 25, 2026, 10:36 PM Larry Garfield <larry@garfieldtech.com> wrote:
>>
>>
>>
>> I agree with Derick.  Using a translator to assist with a language barrier is reasonable, and these days they've all been replaced with LLMs (justified or not).  But I would still prefer to have that explicitly stated, so we know what we're dealing with.  (And can use simpler English in replies, potentially, to aid with understanding.)
>>
>> Beyond that, I want to talk to humans, not slop machines.  If someone's diplomatic skills are not up to the task of discussing on the list without filtering through a chatbot, then they need to work on their diplomatic skills before subjecting us to slop.
>
>
> I have to interfere here again.
>
> I explained a perfectly valid case where llm are used, and they are many orders of magnitude better than translators, even based on llm.
>
> I experience that daily. And the primary reason the personsI work with do it is, guess it, to respect me or others.
>
> To say here, given someone has been directly pointed to via his post, would be extremely bad taken in many cultures (read: insulting, losing face, and they will move away).
>
> As the "official" language is English here and most oss project, I cannot let that go unnoticed.
>
> We must be welcoming for non natice speakers. And for one, I welcome them. And if the price is to have a slightly longer mail at time,  so be it. respect and welcoming must prevail.
>
> --
> Pierre
>
> @pierrejoye

Hi,

Not trying to speak for anyone here, but it seems like the concern
about AI use on the list isn't really about helping with translation
or other genuine values of AI. That has been said explicitly more than
once, and I don't think anyone has actually argued against it.

What has been objected to, though, is a tool authoring the argument
itself, with a person just forwarding it as their own.

Which sounds more like an authorship question rather than an AI one?

If that is roughly right, then I think the problem isn't really about
using AI per se, it's probably more about
engagement/authenticity/authorship. Which is something a policy could
probably be written around.

Additionally, a written policy would make it easier for those who rely
on AI to blend in by drawing the guardrails and explicitly stating
what is and isn't acceptable. Something that would keep the list as
welcoming as it has always been without draining people's energy.

Those were just my two cents...

Regards,
Osama

سپهر محمودی

6 days ago

در تاریخ سه‌شنبه ۲۵ اوت ۲۰۲۶، ۲۲:۲۰ Osama Aldemeery <aldemeery@gmail.com>
نوشت:

> On Tue, Aug 25, 2026 at 7:30 PM Pierre Joye <pierre.php@gmail.com> wrote:
> >
> > hello,
> >
> >
> >
> > On Tue, Aug 25, 2026, 10:36 PM Larry Garfield <larry@garfieldtech.com>
> wrote:
> >>
> >>
> >>
> >> I agree with Derick.  Using a translator to assist with a language
> barrier is reasonable, and these days they've all been replaced with LLMs
> (justified or not).  But I would still prefer to have that explicitly
> stated, so we know what we're dealing with.  (And can use simpler English
> in replies, potentially, to aid with understanding.)
> >>
> >> Beyond that, I want to talk to humans, not slop machines.  If someone's
> diplomatic skills are not up to the task of discussing on the list without
> filtering through a chatbot, then they need to work on their diplomatic
> skills before subjecting us to slop.
> >
> >
> > I have to interfere here again.
> >
> > I explained a perfectly valid case where llm are used, and they are many
> orders of magnitude better than translators, even based on llm.
> >
> > I experience that daily. And the primary reason the personsI work with
> do it is, guess it, to respect me or others.
> >
> > To say here, given someone has been directly pointed to via his post,
> would be extremely bad taken in many cultures (read: insulting, losing
> face, and they will move away).
> >
> > As the "official" language is English here and most oss project, I
> cannot let that go unnoticed.
> >
> > We must be welcoming for non natice speakers. And for one, I welcome
> them. And if the price is to have a slightly longer mail at time,  so be
> it. respect and welcoming must prevail.
> >
> > --
> > Pierre
> >
> > @pierrejoye
>
> Hi,
>
> Not trying to speak for anyone here, but it seems like the concern
> about AI use on the list isn't really about helping with translation
> or other genuine values of AI. That has been said explicitly more than
> once, and I don't think anyone has actually argued against it.
>
> What has been objected to, though, is a tool authoring the argument
> itself, with a person just forwarding it as their own.
>
> Which sounds more like an authorship question rather than an AI one?
>
> If that is roughly right, then I think the problem isn't really about
> using AI per se, it's probably more about
> engagement/authenticity/authorship. Which is something a policy could
> probably be written around.
>
> Additionally, a written policy would make it easier for those who rely
> on AI to blend in by drawing the guardrails and explicitly stating
> what is and isn't acceptable. Something that would keep the list as
> welcoming as it has always been without draining people's energy.
>
> Those were just my two cents...
>
> Regards,
> Osama
>

----------

Hi everyone,

I want to thank Osama and Pierre for their well-articulated points. I
completely agree with the distinction made here: there is a fundamental
difference between using AI as a translation/grammar tool and using it to
generate arguments or act as the author.

As a non-native English speaker, my goal in using translation tools has
always been to ensure my technical points are clear and to communicate
respectfully with the community, never to replace my own thoughts or
technical arguments.

I fully support Osama's suggestion of creating a written policy or
guardrails. Having clear guidelines will help everyone, especially
non-native speakers, know the boundaries while keeping the community
welcoming.

Thank you all for the constructive discussion. I look forward to focusing
our energy back on improving PHP!

Best regards,
Sepehr

Osama Aldemeery

6 days ago

‪> Hi everyone,

Hi Sepehr,

> I want to thank Osama and Pierre for their well-articulated points.

Thank you

> I completely agree with the distinction made here: there is a fundamental difference between using AI as a translation/grammar tool and using it to generate arguments or act as the author.
>
> As a non-native English speaker, my goal in using translation tools has always been to ensure my technical points are clear and to communicate respectfully with the community, never to replace my own thoughts or technical arguments.
>
> I fully support Osama's suggestion of creating a written policy or guardrails. Having clear guidelines will help everyone, especially non-native speakers, know the boundaries while keeping the community welcoming.

I would like to flag one thing about how my earlier message is being
used here, separate from whether I agree with what you've said.
I neither have the authority nor the standing to advise the very
people who maintain this lanaguage on what they should do about a
policy/governance issue, especially on something that could very
easily waste their time and effort.
What I actually offered was an observation about a pattern I'd noticed
recur over the months, a question about it, and my two-cents answer to
that question.

I'd rather that stayed a question people could take or leave, not a
suggestion that is seeking support or trying to avoid opposition.

> Thank you all for the constructive discussion. I look forward to focusing our energy back on improving PHP!

Thank you... And FWIW I do think this is a genuine conversation about
improving PHP.

Best,
Osama

Pierre Joye

6 days ago

Hi Osama,

On Wed, Aug 26, 2026 at 1:48 AM Osama Aldemeery <aldemeery@gmail.com> wrote:

> Not trying to speak for anyone here, but it seems like the concern
> about AI use on the list isn't really about helping with translation
> or other genuine values of AI. That has been said explicitly more than
> once, and I don't think anyone has actually argued against it.
>
> What has been objected to, though, is a tool authoring the argument
> itself, with a person just forwarding it as their own.

I would be interested to know how to differentiate this. It will aslo
pop ups quickly during discussions.

Watermark maybe? ;-)

> If that is roughly right, then I think the problem isn't really about
> using AI per se, it's probably more about
> engagement/authenticity/authorship. Which is something a policy could
> probably be written around.

Please no more policies. It is already too much and one needs a Sr VP
of compliance to parse some of the existing one to make sense of them.

> Additionally, a written policy would make it easier for those who rely
> on AI to blend in by drawing the guardrails and explicitly stating
> what is and isn't acceptable. Something that would keep the list as
> welcoming as it has always been without draining people's energy.

If we start to add even more barriers, we are not welcoming, we are
guarding railing contributions on subjective interpretation of what
one wrote, or how.

Best,

--
Pierre

@pierrejoye

سپهر محمودی

6 days ago

در تاریخ چهارشنبه ۲۶ اوت ۲۰۲۶، ۰۵:۴۹ Pierre Joye <pierre.php@gmail.com>
نوشت:

> Hi Osama,
>
> On Wed, Aug 26, 2026 at 1:48 AM Osama Aldemeery <aldemeery@gmail.com>
> wrote:
>
> > Not trying to speak for anyone here, but it seems like the concern
> > about AI use on the list isn't really about helping with translation
> > or other genuine values of AI. That has been said explicitly more than
> > once, and I don't think anyone has actually argued against it.
> >
> > What has been objected to, though, is a tool authoring the argument
> > itself, with a person just forwarding it as their own.
>
> I would be interested to know how to differentiate this. It will aslo
> pop ups quickly during discussions.
>
> Watermark maybe? ;-)
>
> > If that is roughly right, then I think the problem isn't really about
> > using AI per se, it's probably more about
> > engagement/authenticity/authorship. Which is something a policy could
> > probably be written around.
>
> Please no more policies. It is already too much and one needs a Sr VP
> of compliance to parse some of the existing one to make sense of them.
>
> > Additionally, a written policy would make it easier for those who rely
> > on AI to blend in by drawing the guardrails and explicitly stating
> > what is and isn't acceptable. Something that would keep the list as
> > welcoming as it has always been without draining people's energy.
>
> If we start to add even more barriers, we are not welcoming, we are
> guarding railing contributions on subjective interpretation of what
> one wrote, or how.
>
> Best,
> --
> Pierre
>
> @pierrejoye
>

----------

Hi everyone,

Please, let's bring this discussion to an end.

Continuing this thread is only hurting PHP's reputation and disrespects the
immense hard work of esteemed contributors like Rasmus Lerdorf.

As a newcomer and a contributor specifically to the ext/intl extension, I
believe there's no reason for us to cause unnecessary tension among
ourselves. The current policies are already in place, and we cannot perform
miracles overnight.

Let's focus our energy on building and improving PHP together.

Best regards,
Sepehr

youkidearitai

5 days ago

‪2026年8月26日(水) 20:44 ‫سپهر محمودی‬‎ <sepehrphpr@gmail.com>:‬

>
>
>
> در تاریخ چهارشنبه ۲۶ اوت ۲۰۲۶، ۰۵:۴۹ Pierre Joye <pierre.php@gmail.com> نوشت:
>>
>> Hi Osama,
>>
>> On Wed, Aug 26, 2026 at 1:48 AM Osama Aldemeery <aldemeery@gmail.com> wrote:
>>
>> > Not trying to speak for anyone here, but it seems like the concern
>> > about AI use on the list isn't really about helping with translation
>> > or other genuine values of AI. That has been said explicitly more than
>> > once, and I don't think anyone has actually argued against it.
>> >
>> > What has been objected to, though, is a tool authoring the argument
>> > itself, with a person just forwarding it as their own.
>>
>> I would be interested to know how to differentiate this. It will aslo
>> pop ups quickly during discussions.
>>
>> Watermark maybe? ;-)
>>
>> > If that is roughly right, then I think the problem isn't really about
>> > using AI per se, it's probably more about
>> > engagement/authenticity/authorship. Which is something a policy could
>> > probably be written around.
>>
>> Please no more policies. It is already too much and one needs a Sr VP
>> of compliance to parse some of the existing one to make sense of them.
>>
>> > Additionally, a written policy would make it easier for those who rely
>> > on AI to blend in by drawing the guardrails and explicitly stating
>> > what is and isn't acceptable. Something that would keep the list as
>> > welcoming as it has always been without draining people's energy.
>>
>> If we start to add even more barriers, we are not welcoming, we are
>> guarding railing contributions on subjective interpretation of what
>> one wrote, or how.
>>
>> Best,
>> --
>> Pierre
>>
>> @pierrejoye
>
>
> ----------
>
> Hi everyone,
>
> Please, let's bring this discussion to an end.
>
> Continuing this thread is only hurting PHP's reputation and disrespects the immense hard work of esteemed contributors like Rasmus Lerdorf.
>
> As a newcomer and a contributor specifically to the ext/intl extension, I believe there's no reason for us to cause unnecessary tension among ourselves. The current policies are already in place, and we cannot perform miracles overnight.
>
> Let's focus our energy on building and improving PHP together.
>
> Best regards,
> Sepehr
>

Hi Sepehr

> Continuing this thread is only hurting PHP's reputation and disrespects the immense hard work of esteemed contributors like Rasmus Lerdorf.

That is not for you to say.
To reconfirm your position: your using the AI that trying communicate us.
Your position is tolerated by a small number of non-native English
speakers, likes we.

However, If you undermine that trust, you will immediately lose your position.
If that happens, the community will likely conclude that there is no
point in saying anything to the AI.

Please be careful.
Yuya

--
---------------------------
Yuya Hamada (tekimen)
- https://tekitoh-memdhoi.info
- https://github.com/youkidearitai
-----------------------------

سپهر محمودی

5 days ago

در تاریخ پنجشنبه ۲۷ اوت ۲۰۲۶، ۱۴:۲۳ youkidearitai <youkidearitai@gmail.com>
نوشت:

> ‪2026年8月26日(水) 20:44 ‫سپهر محمودی‬‎ <sepehrphpr@gmail.com>:‬
> >
> >
> >
> > در تاریخ چهارشنبه ۲۶ اوت ۲۰۲۶، ۰۵:۴۹ Pierre Joye <pierre.php@gmail.com>
> نوشت:
> >>
> >> Hi Osama,
> >>
> >> On Wed, Aug 26, 2026 at 1:48 AM Osama Aldemeery <aldemeery@gmail.com>
> wrote:
> >>
> >> > Not trying to speak for anyone here, but it seems like the concern
> >> > about AI use on the list isn't really about helping with translation
> >> > or other genuine values of AI. That has been said explicitly more than
> >> > once, and I don't think anyone has actually argued against it.
> >> >
> >> > What has been objected to, though, is a tool authoring the argument
> >> > itself, with a person just forwarding it as their own.
> >>
> >> I would be interested to know how to differentiate this. It will aslo
> >> pop ups quickly during discussions.
> >>
> >> Watermark maybe? ;-)
> >>
> >> > If that is roughly right, then I think the problem isn't really about
> >> > using AI per se, it's probably more about
> >> > engagement/authenticity/authorship. Which is something a policy could
> >> > probably be written around.
> >>
> >> Please no more policies. It is already too much and one needs a Sr VP
> >> of compliance to parse some of the existing one to make sense of them.
> >>
> >> > Additionally, a written policy would make it easier for those who rely
> >> > on AI to blend in by drawing the guardrails and explicitly stating
> >> > what is and isn't acceptable. Something that would keep the list as
> >> > welcoming as it has always been without draining people's energy.
> >>
> >> If we start to add even more barriers, we are not welcoming, we are
> >> guarding railing contributions on subjective interpretation of what
> >> one wrote, or how.
> >>
> >> Best,
> >> --
> >> Pierre
> >>
> >> @pierrejoye
> >
> >
> > ----------
> >
> > Hi everyone,
> >
> > Please, let's bring this discussion to an end.
> >
> > Continuing this thread is only hurting PHP's reputation and disrespects
> the immense hard work of esteemed contributors like Rasmus Lerdorf.
> >
> > As a newcomer and a contributor specifically to the ext/intl extension,
> I believe there's no reason for us to cause unnecessary tension among
> ourselves. The current policies are already in place, and we cannot perform
> miracles overnight.
> >
> > Let's focus our energy on building and improving PHP together.
> >
> > Best regards,
> > Sepehr
> >
>
> Hi Sepehr
>
> > Continuing this thread is only hurting PHP's reputation and disrespects
> the immense hard work of esteemed contributors like Rasmus Lerdorf.
>
> That is not for you to say.
> To reconfirm your position: your using the AI that trying communicate us.
> Your position is tolerated by a small number of non-native English
> speakers, likes we.
>
> However, If you undermine that trust, you will immediately lose your
> position.
> If that happens, the community will likely conclude that there is no
> point in saying anything to the AI.
>
> Please be careful.
> Yuya
> --
> ---------------------------
> Yuya Hamada (tekimen)
> - https://tekitoh-memdhoi.info
> - https://github.com/youkidearitai
> -----------------------------
>

Dear Yuya—I really like you all, but there is one thing I need to tell you
as a developer.
At first, Mr. Chris snapped at me over this issue, and a few days later,
Mr. Larry was extremely rude—even bringing up the "trash can" analogy—yet I
didn't say a word in return.
I don't use AI—because AI isn't capable of speaking so fluently.
God willing, I am set to become an official contributor to the `intl`
extension; I am currently designing the `array_str_contains` function. I
even work using Gmail on my phone—which is much harder than using the
desktop version—yet you’ve all decided to gang up on poor old me.
Mr. Yuya, I actually included a Japanese example on my RFC
page—specifically for your sake.
I am truly grateful for everyone's comments.
sepehr

weilin-du@qq.com

7 days ago

Hi Christian, sepehr, Yuya and Osama,


We have a AI policy for Github comments already [1]. If you want
to expand it to e-mails and PRs, feel free to open a policy RFC.


Personally, I am okay with AIGC. Fundamentally AI is a tool, and
clearly a useful one. In this thread, I think you treat "use AI or not"
in comments as a "Proof of work" in this era.&nbsp; In most cases, people
using AI to generate text seems to put less effort in it, that's true,
but I don't agree with some of the "progressive" AI policy you've
mentioned in other projects. As a maintainer of the intl extension,
I'd focus more on the code quality itself rather than whether it's
AI-generated code or comments. I frankly don't agree with the
proposal to ban AI in the php internals.


That being said, I feel disgust to communicate with soul-less
contributors who entirely use AI without any critical thinking.
However, it depends. If one can use AI to generate good code
and idea, I am glad to work with these people. I don't see the
benefit to completely ban AIGC in the project.


Also to sepehr: Thank you for your interest in the intl extension!
To get you started: in the extension, some constant can be enums.
And, I would be in favor of appropriate namespaces to avoid conflicts
in the extension's class that make things clearer. Maybe that is a good
RFC idea&nbsp;in 8.7 that needs someone to work on. Also, it is still in
progress to fill in the little blanks between php APIs and ICU APIs. It
might to a good way to start if you want to add new functions/methods.
Not to mention the error state stuff in the extension is, frankly, bad, in
my opinion, and could be optimized since it has caused a lot of very
stupid&nbsp;bugs and make my life harder :( There are still bugs and PRs on
Github you can help with, the list goes on and on and on...


There are tons of things you can do :) This goes to everyone who want to
contribute to the extension.


Cheers,
Weilin Du, weilindu@php.net


[1]: https://github.com/php/php-src/blob/master/CONTRIBUTING.md#llm-usage-in-github-comments&nbsp;

سپهر محمودی

7 days ago

در تاریخ دوشنبه ۲۴ اوت ۲۰۲۶، ۲۳:۳۴ Weilin Du <weilin-du@qq.com> نوشت:

> Hi Christian, sepehr, Yuya and Osama,
>
> We have a AI policy for Github comments already [1]. If you want
> to expand it to e-mails and PRs, feel free to open a policy RFC.
>
> Personally, I am okay with AIGC. Fundamentally AI is a tool, and
> clearly a useful one. In this thread, I think you treat "use AI or not"
> in comments as a "Proof of work" in this era.  In most cases, people
> using AI to generate text seems to put less effort in it, that's true,
> but I don't agree with some of the "progressive" AI policy you've
> mentioned in other projects. As a maintainer of the intl extension,
> I'd focus more on the code quality itself rather than whether it's
> AI-generated code or comments. I frankly don't agree with the
> proposal to ban AI in the php internals.
>
> That being said, I feel disgust to communicate with soul-less
> contributors who entirely use AI without any critical thinking.
> However, it depends. If one can use AI to generate good code
> and idea, I am glad to work with these people. I don't see the
> benefit to completely ban AIGC in the project.
>
> Also to sepehr: Thank you for your interest in the intl extension!
> To get you started: in the extension, some constant can be enums.
> And, I would be in favor of appropriate namespaces to avoid conflicts
> in the extension's class that make things clearer. Maybe that is a good
> RFC idea in 8.7 that needs someone to work on. Also, it is still in
> progress to fill in the little blanks between php APIs and ICU APIs. It
> might to a good way to start if you want to add new functions/methods.
> Not to mention the error state stuff in the extension is, frankly, bad, in
> my opinion, and could be optimized since it has caused a lot of very
> stupid bugs and make my life harder :( There are still bugs and PRs on
> Github you can help with, the list goes on and on and on...
>
> There are tons of things you can do :) This goes to everyone who want to
> contribute to the extension.
>
> Cheers,
> Weilin Du, weilindu@php.net
>
> [1]:
> https://github.com/php/php-src/blob/master/CONTRIBUTING.md#llm-usage-in-github-comments
>
>

---------------


Hi Weilin,

Thank you so much for the warm welcome and for sharing your thoughts!

I completely agree with your perspective on AI. It's a fantastic tool that
boosts productivity, but critical thinking, deeply understanding the logic,
and ensuring code quality must always come first. I always make sure to
fully understand and test every line of code I contribute.

Your suggestions for ext/intl in PHP 8.7—like transitioning constants to
Enums, introducing proper Namespaces, and improving the error state
handling—are absolutely great ideas.

Right now, I am entirely focused on finalizing my RFC for
`array_str_contains` (and related functions) and getting its PR ready.
However, I am very interested in exploring those intl improvements for PHP
8.7 right after I wrap this up.

Thanks again for the support and for pointing me in the right direction!

Best regards,
Sepehr Mahmoudi

سپهر محمودی

7 days ago

در تاریخ سه‌شنبه ۲۵ اوت ۲۰۲۶، ۰۰:۵۳ سپهر محمودی <sepehrphpr@gmail.com> نوشت:

>
>
> در تاریخ دوشنبه ۲۴ اوت ۲۰۲۶، ۲۳:۳۴ Weilin Du <weilin-du@qq.com> نوشت:
>
>> Hi Christian, sepehr, Yuya and Osama,
>>
>> We have a AI policy for Github comments already [1]. If you want
>> to expand it to e-mails and PRs, feel free to open a policy RFC.
>>
>> Personally, I am okay with AIGC. Fundamentally AI is a tool, and
>> clearly a useful one. In this thread, I think you treat "use AI or not"
>> in comments as a "Proof of work" in this era.  In most cases, people
>> using AI to generate text seems to put less effort in it, that's true,
>> but I don't agree with some of the "progressive" AI policy you've
>> mentioned in other projects. As a maintainer of the intl extension,
>> I'd focus more on the code quality itself rather than whether it's
>> AI-generated code or comments. I frankly don't agree with the
>> proposal to ban AI in the php internals.
>>
>> That being said, I feel disgust to communicate with soul-less
>> contributors who entirely use AI without any critical thinking.
>> However, it depends. If one can use AI to generate good code
>> and idea, I am glad to work with these people. I don't see the
>> benefit to completely ban AIGC in the project.
>>
>> Also to sepehr: Thank you for your interest in the intl extension!
>> To get you started: in the extension, some constant can be enums.
>> And, I would be in favor of appropriate namespaces to avoid conflicts
>> in the extension's class that make things clearer. Maybe that is a good
>> RFC idea in 8.7 that needs someone to work on. Also, it is still in
>> progress to fill in the little blanks between php APIs and ICU APIs. It
>> might to a good way to start if you want to add new functions/methods.
>> Not to mention the error state stuff in the extension is, frankly, bad, in
>> my opinion, and could be optimized since it has caused a lot of very
>> stupid bugs and make my life harder :( There are still bugs and PRs on
>> Github you can help with, the list goes on and on and on...
>>
>> There are tons of things you can do :) This goes to everyone who want to
>> contribute to the extension.
>>
>> Cheers,
>> Weilin Du, weilindu@php.net
>>
>> [1]:
>> https://github.com/php/php-src/blob/master/CONTRIBUTING.md#llm-usage-in-github-comments
>>
>>
>
> ---------------
>
>
> Hi Weilin,
>
> Thank you so much for the warm welcome and for sharing your thoughts!
>
> I completely agree with your perspective on AI. It's a fantastic tool that
> boosts productivity, but critical thinking, deeply understanding the logic,
> and ensuring code quality must always come first. I always make sure to
> fully understand and test every line of code I contribute.
>
> Your suggestions for ext/intl in PHP 8.7—like transitioning constants to
> Enums, introducing proper Namespaces, and improving the error state
> handling—are absolutely great ideas.
>
> Right now, I am entirely focused on finalizing my RFC for
> `array_str_contains` (and related functions) and getting its PR ready.
> However, I am very interested in exploring those intl improvements for PHP
> 8.7 right after I wrap this up.
>
> Thanks again for the support and for pointing me in the right direction!
>
> Best regards,
> Sepehr Mahmoudi
>

----------

Hi everyone,

Regarding the topic of AI: In my opinion, AI tends to write text in a very
literary and formal manner, which makes it useful for that specific
purpose. However, if one wants to use it for coding—which I personally do
not do—it requires giving it exact, step-by-step instructions to get the
desired result.

While some developers might choose to use AI for their code, my personal
approach is different. I prefer to work entirely independently and rely on
my own skills for both coding and communications.

Hopefully, after finalizing my current RFC (`array_str_contains`), I will
begin my work on the `intl` extension.

Best regards,
Sepehr

Marco Pivetta

7 days ago

Hey Christian,

On Sun, 23 Aug 2026 at 15:21, Christian Schneider <cschneid@cschneid.com>
wrote:

> In the thead "[PHP-DEV] [RFC] [Discussion] array_match" I got a reply
> starting with "Thanks for your feedback and for raising these valid points"
> and then sort of reiterating things I already addressed ("requires adding
> delimiters and wrapping the needle in `preg_quote()`") which leads me to
> believe that the response was AI generated. I might be wrong in this
> specific case but it got me thinking.
>
> I'm not sure I want to spend time arguing with AIs and was considering
> replying with "AI;DR" (AI; didn't reply) but I realize that this can be
> considered rude.
>
> Just to be clear: I realize that AIs help a lot for non-native speakers to
> express themselves but when I get the feeling that the whole argument was
> generated and not only translated then I get a bit defensive about my
> personal time and energy.
>
> As I'm expecting more and more generated mails on this list: What is the
> feeling on the mailing list about how to handle this?
>
> Regards,
> - Chris
>

Just my personal data point here: the responses to this thread gave me
nausea.

If people want to act as meat proxy for their glorified Markov chains, they
should at least be extremely careful about the output, and avoid the
Vogon-speak I've seen here.

The Rust project has an **EXCELLENT** policy on LLM usage:
https://github.com/rust-lang/rust-forge/blob/382ef09a0aae87b149cdcd4866cfd72b6a176a1e/src/policies/llm-usage.md#-banned

The TLDR is: humans shall not suffer Vogon-speak generated by a machine.

Marco Pivetta

https://mastodon.social/@ocramius

https://ocramius.github.io/

Pierre Joye

7 days ago

hello Christian,

On Sun, Aug 23, 2026, 8:24 PM Christian Schneider <cschneid@cschneid.com>
wrote:

As I'm expecting more and more generated mails on this list: What is the

> feeling on the mailing list about how to handle this?
>

I am usually ok with it. I am not a fan however I totally understand why a
few use it. And to see a forgotten line in the copy/paste is not needed to
see it.

Many are not native English speakers, or not western languages (double
barriers). Add the possible pressure created to post a rfc, a patch etc to
internals and I would use it as well to make sure my message comes here
well and correctly formulated.

cheers,
Pierre

--
Pierre

@pierrejoye

سپهر محمودی

7 days ago

در تاریخ سه‌شنبه ۲۵ اوت ۲۰۲۶، ۱۷:۲۵ Pierre Joye <pierre.php@gmail.com> نوشت:

> hello Christian,
>
>
>
> On Sun, Aug 23, 2026, 8:24 PM Christian Schneider <cschneid@cschneid.com>
> wrote:
>
> As I'm expecting more and more generated mails on this list: What is the
>> feeling on the mailing list about how to handle this?
>>
>
> I am usually ok with it. I am not a fan however I totally understand why a
> few use it. And to see a forgotten line in the copy/paste is not needed to
> see it.
>
> Many are not native English speakers, or not western languages (double
> barriers). Add the possible pressure created to post a rfc, a patch etc to
> internals and I would use it as well to make sure my message comes here
> well and correctly formulated.
>
> cheers,
> Pierre
>
> --
> Pierre
>
> @pierrejoye
>

-----------


Hi Pierre,

Thanks for jumping in with such a great email. You really nailed exactly
what it's like for non-native English speakers.

I really appreciate you bringing some much-needed perspective to the thread
and helping keep things constructive.

Cheers,
Sepehr

Juris Evertovskis

6 days ago

> -----Original Message-----
> From: Christian Schneider <cschneid@cschneid.com>
> Sent: Sunday, August 23, 2026 4:21 PM
> To: php internals <internals@lists.php.net>
> Subject: [PHP-DEV] AI generated mails
>
> I'm not sure I want to spend time arguing with AIs and was considering
> replying with "AI;DR" (AI; didn't reply) but I realize that this can be
> considered rude.

Helo Mr Chris and Mr Mailing List,

I think that would be too rejecty without having rules/guidelines in place.

It would be best to have some guidelines (on the same level as "u must
bottompost") so people can read them before posting. And, on par with bottom
posting reminders, it would be fair to provide an eager newcomer with a
guiding reply, something along the lines:

> It is expected to write the texts yourself on this mailing list. Please do
> not write or rephrase yourself using an LLM.
> If English is the problem, do not worry — there is no requirement to have
> perfect English as long as you can understand what is going on. A lot of
> productive contributors here are more fluent in C and PHP languages than
> they are in English.
> And if you must, use a deterministic translator that will represent what

you

> intend to say, instead of something that the LLM will say.

Probably a native speaker can write a better and more concise version of the
above :))

> Just to be clear: I realize that AIs help a lot for non-native speakers to
> express themselves

I don't buy that. There's nothing about being native or non-native in having
AI speak instead of you, so... [translation below]

Lai pierādītu, ka ģeneratīvais MI nav nepieciešams, lai izteiktos, tālāko
rakstīšu latviski un tulkošu ar Google tulkotāju.

Jā, Pjēram pilnīga taisnība, ka mums jābūt pretimnākošiem pret dalībniekiem,
kam angļu valoda nav pirmā. Bet tāpēc jau nav jāprasa LLM čatam rakstīt savā
vietā. Protams, ka Google tulkotāja iztulkotais teksts nebūs perfekts un
noteikti toņa ziņā nebūs identisks tam, kā šis teksts skan latviski. Taču
tas būs daudz tuvāks oriģinālam nekā tad, ja es pateiktu savu domu ChatGPT
vai Gemini un lūgtu viņam noformulēt to angliski.

Vēl trakāk — šī teksta angliskā versija būs daudz tuvāka latviskajam
oriģinālam nekā tad, ja vienkārši palūgtu LLM rīkam noformulēt šo domu pat
nemainot valodu.
----------------
To prove that generative AI is not necessary to express myself, I will write
the following in Latvian and translate it with Google Translate.

Yes, Pierre is absolutely right that we should be accommodating to
participants whose first language is not English. But that does not mean
that we should ask the LLM chat to write for us. Of course, the text
translated by Google Translate will not be perfect and certainly will not be
identical in tone to how this text sounds in Latvian. But it will be much
closer to the original than if I told ChatGPT or Gemini my idea and asked it
to formulate it in English.

Even worse, the English version of this text will be much closer to the
original than if I simply asked the LLM tool to formulate this idea without
even changing the language.

BR,
Juris