A conversation about AI in PCRE2

[Nicholas Wilson]

Hello contributors and community,

I would welcome hearing from any of you how you feel about AI and PCRE2.

As my own personal context, I have been a bit of a late adopter, using it in my job for about a year. Because I work with a very large legacy codebase at Microsoft, my particular team hasn't had much satisfaction with the code written by models. However, I use it constantly to review as I go, do broad code-wise searches, prove/disprove invariants, and for large-but-mechanical tasks.

Just this summer, GPT 5.6 Sol (and potentially Opus 5) have finally become models good enough to actually start writing code for me. It will take a while for me to adjust my workflows, and I am fussy and will probably tweak and rewrite much of the output for the time being.

Still, I think everyone with a professional software job is probably in the same situation: hard to stop using it at this point.

As for PCRE2, we haven't discussed what guidelines or philosophy we have about that.

I do enjoy the 'craft' of software, typing and reasoning about the details myself. I will absolutely not use any code in PCRE2 that I do not understand very carefully myself! (Although for example Salvatorre says Control the ideas, not the code - <antirez>).

PCRE2 is partly a hobby project, but it is also used in serious software.

In my personal life, I spend time with my children, my family, and enjoy other hobbies as well as PCRE2. I will never have the time to do all of the things I want to do, unless I make a big effort to reallocate that time.

With AI, I can see quite rapid development and improvement. I have many plans and ideas myself. Previously, large-scale change would have been impossible for me. In an hour or two a week, I can't make changes across all of PCRE (eg updating the UTF-8 handling).

I am happy to set out ideas for what I would like to do with PCRE2.

First however, I would welcome views and feelings from the community.

Do you have principles or guidelines you think we should adopt?

All the best,

Nick

[Philip Hazel]

I have nothing to contribute to this discussion, but I would like to take this opportunity to publicly thank Nick for all the work he has done and is doing in maintaining and developing PCRE2. In particular, all the infrastructure updates are something I could never have attempted. I am really lucky to have been able to hand over my project into such good hands.

Regards,

Philip

--
You received this message because you are subscribed to the Google Groups "PCRE2 discussion list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pcre2-dev+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pcre2-dev/d10d4d42-8aaa-470a-91a6-1ad2f6353ae3n%40googlegroups.com.

[Matthew Vernon]

Hi,

On 24/08/2026 14:38, Nicholas Wilson wrote:

> I would welcome hearing from any of you how you feel about AI and PCRE2.

I am, broadly, opposed in principle to the use of generative AI; I wrote
a little about this on the debian-vote mailing list recently[0]. I have
secondary concerns about how well we understand and review code produced
agentically, and how maintainable it will be in the medium-long term,
particularly in a library like PCRE2 which is both "infrastructure",
quite complex, and often on security boundaries.

[while I'm writing - how are you getting on with a 10.48 RC?]

Regards,

Matthew

[0] https://lists.debian.org/debian-vote/2026/02/msg00060.html

[Nicholas Wilson]

Thank you for your kind words Philip. 

Matthew, I have just read your post and many others from that thread. I am glad to see that Debian is a safe place for these views to be heard. There is a real danger in the current culture of software around AI to shut out discussions. I also particularly admire your willingness to let your moral views lead you, rather than give in to expediency. 

I sympathise with several of your points, although on the whole I feel I have chosen to use AI but advocate for doing so thoughtfully. 

It would be excellent in my opinion if Debian, RedHat, Apache foundation, Linux, and others, could use their combined influence to liase with Anthropic and OpenAI to work with them on their ethics. It may be too late to stop the industry-wide use of AI, but there are years ahead of us while the landscape is in flux for collective steering. Similarly, Azure could be pressured into more disclosure, for example, APIs to report on emissions from their tokens (kWh and % renewables spot) to enable supply chain reporting.

All the best,

Nick

[Zoltan Herczeg]

Hi,

what are the legal implications of using an AI? Who owns the code? Can
an AI accept the license terms?

Regards,
Zoltan

> --
> You received this message because you are subscribed to the Google Groups "PCRE2 discussion list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to pcre2-dev+...@googlegroups.com.

> To view this discussion visit https://groups.google.com/d/msgid/pcre2-dev/89228b8b-4d07-4f64-8029-afafc92b817en%40googlegroups.com.