www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Use of IA for PR - my POV

reply user1234 <user1234 12.de> writes:
For some reasons I monitor a full hand of programming languages 
repos.This is interesting because I can compare to others.

One tendency I have noticed recently in the D world is one guy 
that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
masters the tools.

One problem I'd like to mention is that if for "this" or "that" 
AI may help, you have the risk of not understanding the codebase 
anymore, after merging 100 PR, let's say, the knowledge will be 
lost or costly to re-acquire.

The paradox is however that often human-generated PR are just 
patches. The compiler is a huge patchwork, even if there's many 
sub programs ( visit this or that ).

So you see it's a bit the same. People come and go, newcomers 
have to understand what people who left did. This is what matters.
Feb 09
next sibling parent reply Dejan Lekic <dejan.lekic gmail.com> writes:
On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 For some reasons I monitor a full hand of programming languages 
 repos.This is interesting because I can compare to others.

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.

 One problem I'd like to mention is that if for "this" or "that" 
 AI may help, you have the risk of not understanding the 
 codebase anymore, after merging 100 PR, let's say, the 
 knowledge will be lost or costly to re-acquire.


False claims from people who hide behind random nicknames are 
typically completely ignored by the community. Rightfully so. 
Cybershadow is a well-respected member of this community, 
contributing to the D ecosystem probably close to 20 years if I 
remember well. He also happens to be the guy who wrote the very 
software you used to write your message, in D ofc.

Rest assured that IF he used AI to do his PRs, he _DOES_ 
understand that code.
Feb 10
next sibling parent Serg Gini <kornburn yandex.ru> writes:
On Tuesday, 10 February 2026 at 12:43:49 UTC, Dejan Lekic wrote:

 Rest assured that IF he used AI to do his PRs, he _DOES_ 
 understand that code.


We can only hope this is the case, but I would say we still need 
to be careful.
At the minimum policy we should ask everyone (even long time 
contributors) explicitly and clearly adding in PR information 
that they were using AI for the PR.

This should be applied for Dub packages as well. I think it would 
be a nice and responsible position of package authors to the 
community to let their users know this..
I already identified some packages in Dub registry that 99% were 
fully AI generated.

And as DMD story with ai.d showed - nobody could be strong enough 
versus AI code ..
Feb 10
prev sibling next sibling parent reply user1234 <user1234 12.de> writes:
On Tuesday, 10 February 2026 at 12:43:49 UTC, Dejan Lekic wrote:

 On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 For some reasons I monitor a full hand of programming 
 languages repos.This is interesting because I can compare to 
 others.

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.

 One problem I'd like to mention is that if for "this" or 
 "that" AI may help, you have the risk of not understanding the 
 codebase anymore, after merging 100 PR, let's say, the 
 knowledge will be lost or costly to re-acquire.


 False claims from people who hide behind random nicknames are 
 typically completely ignored by the community.


Coincidence but I think there's another one since today: 
https://github.com/dlang/dmd/pull/22550

I see CyberShadow PR based on Claude getting closed. Curious 
about what will happen for this one.
Feb 10
parent monkyyy <crazymonkyyy gmail.com> writes:
On Tuesday, 10 February 2026 at 14:14:00 UTC, user1234 wrote:

 
 https://github.com/dlang/dmd/pull/22550
 (ai slop)



- Ambiguous template instantiation without parens crashes 
compiler


Is there a real example of this? I cant imagine how a template 
init can be ambiguous
Feb 10
prev sibling parent reply Serg Gini <kornburn yandex.ru> writes:
On Tuesday, 10 February 2026 at 12:43:49 UTC, Dejan Lekic wrote:

 Rest assured that IF he used AI to do his PRs, he _DOES_ 
 understand that code.


Literally next PR )))

"What do you think about this one? I don't fully understand it, 
but it's green and the diff is small 😛"
Feb 10
parent Dejan Lekic <dejan.lekic gmail.com> writes:
On Tuesday, 10 February 2026 at 14:28:46 UTC, Serg Gini wrote:

 On Tuesday, 10 February 2026 at 12:43:49 UTC, Dejan Lekic wrote:

 Rest assured that IF he used AI to do his PRs, he _DOES_ 
 understand that code.


 Literally next PR )))

 "What do you think about this one? I don't fully understand it, 
 but it's green and the diff is small 😛"


He will understand it, have faith.
Feb 10
prev sibling next sibling parent reply Vladimir Panteleev <thecybershadow.lists gmail.com> writes:
On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.


I guess I could post a few thoughts about AI / LLMs here if 
people are interested.

It's not a new thing that I've been interested in the idea of 
offloading menial work onto machines.  DustMite was my first big 
project in that vein - you define an oracle, then drop your xMLOC 
codebase on it and go enjoy your weekend; then Digger to help 
automate bisecting regressions with our multi-repo setup, so LLMs 
are kind of in that vein, if applied properly.

The LLMs themselves were pretty much useless toys for a long time 
when it came to writing code, and the vast majority of them still 
are. I think even what you get today on e.g. chatgpt.com is going 
to be underwhelming from many perspectives. However, it does seem 
like there was a huge jump last year with Opus 4.5.

I've been experimenting with LLMs generating D and other code 
throughout the last year.  Just last August I was playing around 
with the best model at the time - the results were, frankly, 
depressing. I think I spent $200 in tokens for a development 
process that I could have done much faster, prettier, more 
correct, etc. At that point, it was clear that there was nothing 
to be gained from agentic coding, at least for what I was doing.

Then, Opus 4.5 came out. I'm not sure if it really was an 
objectively major breakthrough in capabilities, or if it merely 
crossed some threshold that would quantify it as such, or if that 
was just my perception, but for me it was the first model that 
seemed actually ... useful.

- It could write non-trivial code - entire multi-module programs 
- that actually worked on first try.
- It no longer regularly made stupid mistakes that no human would 
ever do.
- It wrote correct D! Without hallucinating features!

But what shocked me the most:

- It knew about my personal D libraries! Somehow, my personal ae 
D library was in its training data set, with sufficient coverage 
that it could even use some parts of it blind!
- Sometimes, it even wrote better D code than me! It would use 
patterns that I was not aware of or thought of!

To me, this was mind-blowing, and turned my whole programming 
life upside down.

Since then, I decided to do an experiment: Could I just use it 
for everything? Just stop writing code, and use it to do the code 
writing?

Would this make me more productive or less? Would I eventually 
forget how to code? Would I eventually get buried by the big pile 
of slop and broken code that I don't understand?

I didn't know the answers but I was really fascinated to try and 
find out. So, I got a $200/mo Claude Code Max subscription and 
set out with that self-imposed constraint.

Here's what I can tell you so far:

- I'm probably not as good as I was at hammering out code with a 
keyboard as I was last year. But, I do feel like I'm now much 
better at code review, multitasking, and task switching. Since 
the whole idea of AI is to make the bot work on your behalf, you 
can multi-track several projects (or multiple aspects of a 
project), or simply enjoy your hobby while checking in on the bot 
every half hour. Like with DustMite, if you use AI but then stare 
at the screen while it's working, you're Doing It Wrong.
- The bot obviously still makes mistakes. But the mistakes it 
makes are different than the kind of mistakes a human would make: 
no typos, no copy-paste errors, no "I forgot to add this one 
line", no "I used the variable `foo` from the argument list 
instead of the variable `foo` from the local scope".
- On the other hand, the bot is terrible at designing. The APIs 
are bad, the patterns are bad, the structure is bad. You still 
need to think ahead about what you want to build and what shape 
each part should be in. However, you now have a lot more 
cognitive bandwidth to focus on this exclusively.
- Obviously you do need to read and understand the code it 
writes...
- ...unless it's for one-off throwaway scripts, which are now 
really really easy to produce! You can script anything easily 
with zero investment, which is a big help sometimes.
- Other things it's good at:
   - Bug hunting - drop a test case on it, come back half an hour 
later, and it will likely have found the root cause (and maybe 
even an initial patch for it).
   - Code research - have a technical question about a project and 
want a precise answer? `git clone` the GitHub repo, run the 
agent, and ask your question - you'll get an answer with 
citations to exact line numbers.
   - Speculative refactoring - have a complicated code base but 
don't want to invest the time in a refactoring that may make the 
code simpler or it may make it an even bigger mess? The bots are 
very good at mechanical code transformations, so you can give it 
10 refactoring ideas and just leave it overnight.
   - Writing test cases, but everyone knows this one already.
- In terms of getting things done, I do find myself to be more a 
lot more productive! Certainly not in terms of time, but 
definitely creative energy. I've picked up and even wrapped up a 
lot of projects from my backlog. I do miss some bugs in review 
(or sometimes am just too lazy to review the code), so the output 
quality is maybe not the same as what I would have churned out by 
hand, but I'll definitely take a 20% quality hit for a 500% 
productivity gain.

On contributing to D:

I think so far I used Claude to write patches for Phobos, 
Druntime, and DMD. In order:

- Phobos: For me these are very easy to review. I'm confident in 
their quality, so it's just a time saver.
- Druntime: These have been mainly translations of C headers to 
D. LLMs are good at these, so the main thing to watch out for is 
that the translation follows our conventions.

And then there's the compiler, DMD.

So, here's the thing. Maybe my perspective is off, but my point 
of view is: in order to understand and be able to meaningfully 
review patches to the compiler, you must be a compiler developer. 
And, you do not simply become a compiler developer.

As much as I wish I could understand and help out with all parts 
of D, I need to pick my battles. In my mind, D compiler hackers 
are the most elite of the elite D developers. I bow to them and 
plead for their mercy as they consider my bug reports and patches.

This puts me in a difficult situation every time I run into a 
blocking compiler bug. I could:

1. Reduce the bug to a test case, file an issue, and watch as 
likely absolutely nothing happens for years (blocker or not, 
regression or not). Understandable, since compiler bugs are hard, 
compiler development is hard, and nothing in life is free.
2. Give up on some of my personal projects and invest into 
becoming a D compiler developer instead.
3. Tuck my tail and try to work around it in my code base, giving 
up on my perfect envisioned design.
4. [NEW!] Ask the bot to draft a patch, which it often ultimately 
succeeds at doing (at least to the point of getting the test 
suite to pass). Now, instead of filing a bug report, I can file a 
bug report with a machine-generated patch attached (in the form 
of a pull request), which might be total garbage - but at least 
it starts a discussion! I'm not sure how the compiler hackers 
feel about this, though, but I've always tried to be up-front 
about the provenance of the patches and so far I have not been 
asked to stop.

What should we do about this? Should we use it more? Should we 
use it less? I don't know. There's definitely valid reasons - 
ethical, practical, legal, financial - to avoid using it, as have 
been mentioned here and elsewhere. But it also seems genuinely 
useful in at least some situations, and no one knows what the 
future holds.

Anyway, one point I wanted to make is that if we're talking about 
the quality/implications/etc. of AI, we should definitely be 
clear about which specific model we're talking about. Things have 
been improving very rapidly and there's a lot of variation in 
what you might experience recently or even today.
Feb 10
next sibling parent user1234 <user1234 12.de> writes:
On Tuesday, 10 February 2026 at 16:14:03 UTC, Vladimir Panteleev 
wrote:

 On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.


 I guess I could post a few thoughts about AI / LLMs here if 
 people are interested.
 [...]


thanks much for that reply.
Feb 10
prev sibling next sibling parent reply matheus <matheus gmail.com> writes:
On Tuesday, 10 February 2026 at 16:14:03 UTC, Vladimir Panteleev 
wrote:

 On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.


 I guess I could post a few thoughts about AI / LLMs here if 
 people are interested...


Interesting, since I'm not using AI I'd like to know, in this 
case you have LLM locally and you point to D source folder and It 
learns from that database and do everything from there?

I think this would be a nice topic/video to be made so to attract 
people, since D suffers from content lately. And show how do 
you're doing PR currently, maybe even guys like me who will go 
deep to help would try it.

*** Now On the subject ***

I remember a DConf with Scott Meyers where they talked about the 
language that would Kill "C", and I'm starting to think that AI 
could kill any language. I read a topic about this on Spectrum, 
and just as example, where I work (One of the biggest care 
provider in my country), they showed us a port of one module 
written over the past 20 years in a new language using AI in just 
a couple of hours, it was modernized and everything else through 
AI.

So I wonder if usually programming languages have restrictions to 
ensure bad code don't mess with anything, but on the other hand 
AI keep getting better and learns how to avoid bad code, what's 
the point of having all these languages? Or in fact, could AI 
write a better programming language by itself?

Matheus.
Feb 10
next sibling parent reply Vladimir Panteleev <thecybershadow.lists gmail.com> writes:
On Tuesday, 10 February 2026 at 17:38:40 UTC, matheus wrote:

 Interesting, since I'm not using AI I'd like to know, in this 
 case you have LLM locally and you point to D source folder and 
 It learns from that database and do everything from there?


The main way I use LLMs is with Claude Code. Here's how it works:

1. You open the directory with your project in a terminal
2. You run `claude`
3. This opens a TUI that looks like a chat interface. You type 
your question or request for what you want the bot to do.
4. The bot looks at your code. If it's too big to fit into its 
context (a limited window of how much it can see at a time), it 
will search just for relevant bits.
5. If the problem is big, it will first write a plan for how it 
aims to accomplish its goal, for you to read and approve.
6. It does the thing. It can edit files and run commands in order 
to run your test suite (or at least check that the code 
compiles). By default it will ask before every edit or command. 
Many people run it in a sandbox and disable the prompts, so that 
it can work by itself but still doesn't accidentally delete your 
entire computer.
7. Sometimes the bot can automatically write down what it learned 
in a memory file. It will read this file automatically the next 
time you ask it to do something in that project. There isn't 
really a lot of "learning" other than something like this.

Before/aside that, I also have a spare GPU which I'm using to run 
an autocomplete model. It's nice when writing code by hand. For 
that I use https://github.com/CyberShadow/company-llama + 
llama.cpp.


 So I wonder if usually programming languages have restrictions 
 to ensure bad code don't mess with anything, but on the other 
 hand AI keep getting better and learns how to avoid bad code, 
 what's the point of having all these languages? Or in fact, 
 could AI write a better programming language by itself?


Agentic coding actually works better the stricter the language! 
This is because then the compiler can check if the code is 
correct immediately, and if it's not, the agent sees the error 
right away and can fix it before stopping. So, I think we will 
see more strictly typed languages or languages with built-in 
theorem proving become more popular. There are often too 
frustrating or time consuming for humans to use for every day 
programming, but it doesn't matter when the code is being written 
by AI.

I am seeing this too with testing and Nix. Writing integration 
tests with Nix is usually a lot of work, but once it's written 
then it's rock-solid proof that your thing works and that 
everyone can verify that it works. You can even script entire VMs 
that can run any software for integration tests, and these VM 
tests run without any problems on any Linux machine including 
GitHub Actions. So, I've since been adding Nix based integration 
tests to all my projects (including this forum, which now has 
Nix/Playwright based end-to-end tests).
Feb 10
next sibling parent matheus <matheus gmail.com> writes:
On Tuesday, 10 February 2026 at 17:54:29 UTC, Vladimir Panteleev 
wrote:

 On Tuesday, 10 February 2026 at 17:38:40 UTC, matheus wrote:

 Interesting, since I'm not using AI I'd like to know, in this 
 case you have LLM locally and you point to D source folder and 
 It learns from that database and do everything from there?


 The main way I use LLMs is with Claude Code. Here's how it 
 works:
 ...


Well first of all thanks for sharing this info, It's interesting 
and some of the things are the way I had imagined.

I read somewhere that Stack Exchange programming questions is 
only 22% of what it had 2 years ago[1], I think development with 
LLM will be the norm now.

I tried a bit of AI sometime ago and it wasn't very pleasant 
experience, in fact it would be faster writing myself than fixing 
AI terrible code, but for what people are talking these days it 
has evolved a lot.

Finally and again thanks for sharing how you are using it,

Matheus.

[1] - https://spectrum.ieee.org/top-programming-languages-2025
Feb 10
prev sibling parent reply user1234 <user1234 12.de> writes:
On Tuesday, 10 February 2026 at 17:54:29 UTC, Vladimir Panteleev 
wrote:

 On Tuesday, 10 February 2026 at 17:38:40 UTC, matheus wrote:

 Interesting, since I'm not using AI I'd like to know, in this 
 case you have LLM locally and you point to D source folder and 
 It learns from that database and do everything from there?


 The main way I use LLMs is with Claude Code. Here's how it 
 works:

 1. You open the directory with your project in a terminal
 2. You run `claude`
 3. This opens a TUI that looks like a chat interface. You type 
 your question or request for what you want the bot to do.
 4. The bot looks at your code. If it's too big to fit into its 
 context (a limited window of how much it can see at a time), it 
 will search just for relevant bits.
 5. If the problem is big, it will first write a plan for how it 
 aims to accomplish its goal, for you to read and approve.
 6. It does the thing. It can edit files and run commands in 
 order to run your test suite (or at least check that the code 
 compiles). By default it will ask before every edit or command. 
 Many people run it in a sandbox and disable the prompts, so 
 that it can work by itself but still doesn't accidentally 
 delete your entire computer.
 7. Sometimes the bot can automatically write down what it 
 learned in a memory file. It will read this file automatically 
 the next time you ask it to do something in that project. There 
 isn't really a lot of "learning" other than something like this.

 Before/aside that, I also have a spare GPU which I'm using to 
 run an autocomplete model. It's nice when writing code by hand. 
 For that I use https://github.com/CyberShadow/company-llama + 
 llama.cpp.


 So I wonder if usually programming languages have restrictions 
 to ensure bad code don't mess with anything, but on the other 
 hand AI keep getting better and learns how to avoid bad code, 
 what's the point of having all these languages? Or in fact, 
 could AI write a better programming language by itself?


 Agentic coding actually works better the stricter the language! 
 This is because then the compiler can check if the code is 
 correct immediately, and if it's not, the agent sees the error 
 right away and can fix it before stopping. So, I think we will 
 see more strictly typed languages or languages with built-in 
 theorem proving become more popular. There are often too 
 frustrating or time consuming for humans to use for every day 
 programming, but it doesn't matter when the code is being 
 written by AI.

 I am seeing this too with testing and Nix. Writing integration 
 tests with Nix is usually a lot of work, but once it's written 
 then it's rock-solid proof that your thing works and that 
 everyone can verify that it works. You can even script entire 
 VMs that can run any software for integration tests, and these 
 VM tests run without any problems on any Linux machine 
 including GitHub Actions. So, I've since been adding Nix based 
 integration tests to all my projects (including this forum, 
 which now has Nix/Playwright based end-to-end tests).


I feel so old-fashioned when I read this. The worst is that I've 
been warned, pre-Covid time (so 2019), while chating, that this 
will the next big thing. You seem to have the method. I've heard 
recently that apparently another notable D user, "Feep", is 
heavily using AI agents.

I'm so done.
Feb 10
next sibling parent reply monkyyy <crazymonkyyy gmail.com> writes:
On Tuesday, 10 February 2026 at 20:16:06 UTC, user1234 wrote:

 I feel so old-fashioned when I read this. The worst is that 
 I've been warned, pre-Covid time (so 2019), while chating, that 
 this will the next big thing. You seem to have the method. I've 
 heard recently that apparently another notable D user, "Feep", 
 is heavily using AI agents.

 I'm so done.


I dont think ai will ever replace writing new abstractions. Its 
like an oo-ide writing 30 lines of getters and setters instantly. 
Thats great, such a  marvel of engineering... it shouldn't 
actually been a problem to solve.

99% of lines of code will be ai. But, what does that actually 
mean.
Feb 10
parent "H. S. Teoh" <hsteoh qfbox.info> writes:
On Tue, Feb 10, 2026 at 09:06:13PM +0000, monkyyy via Digitalmars-d wrote:

 On Tuesday, 10 February 2026 at 20:16:06 UTC, user1234 wrote:

 I feel so old-fashioned when I read this. The worst is that I've
 been warned, pre-Covid time (so 2019), while chating, that this will
 the next big thing. You seem to have the method. I've heard recently
 that apparently another notable D user, "Feep", is heavily using AI
 agents.
 
 I'm so done.


 
 I dont think ai will ever replace writing new abstractions. Its like
 an oo-ide writing 30 lines of getters and setters instantly. Thats
 great, such a  marvel of engineering... it shouldn't actually been a
 problem to solve.
 
 99% of lines of code will be ai. But, what does that actually mean.


To pull out a Walter quote:

	I've been around long enough to have seen an endless parade of
	magic new techniques du jour, most of which purport to remove
	the necessity of thought about your programming problem.  In the
	end they wind up contributing one or two pieces to the
	collective wisdom, and fade away in the rearview mirror.
	-- Walter Bright

IMNSHO, if something can be automatically generated by AI, then it's not
worth writing and shouldn't have been necessary to write in the first
place.  What's worth writing is what requires actual thought, actually
solving a problem, not just regurgitating past solutions to
previously-solved problems.  If AI can automate those parts away, so
much the better, I say.  Spare me the tedium, give me more time/energy
to focus on what actually needs solving.


T

-- 
What's at the bottom of the Bermuda Triangle? A wreck tangle.
Feb 10
prev sibling parent FeepingCreature <feepingcreature gmail.com> writes:
On Tuesday, 10 February 2026 at 20:16:06 UTC, user1234 wrote:

 I feel so old-fashioned when I read this. The worst is that 
 I've been warned, pre-Covid time (so 2019), while chating, that 
 this will the next big thing. You seem to have the method. I've 
 heard recently that apparently another notable D user, "Feep", 
 is heavily using AI agents.

 I'm so done.


hi! yep, I use AI heavily. I'm a doomer as well fwiw, so I'm by 
no means uncritical about AI. But programming has *always* been 
the art of getting done as much as possible with the least 
possible effort. AI simply continues the trend.

Fwiw, despite using AI very heavily in private, my D code at work 
is still largely written manually. The more local context and 
insight is required, the more current AI struggles. Maybe next 
year we'll have local online learning... :)

(Also yes, it's a trip that Claude knows boilerplate/serialized. 
The case for open-sourcing internal corporate code has never been 
stronger.)
Feb 11
prev sibling next sibling parent Julian Fondren <julian.fondren gmail.com> writes:
On Tuesday, 10 February 2026 at 17:38:40 UTC, matheus wrote:

 I remember a DConf with Scott Meyers where they talked about 
 the language that would Kill "C", and I'm starting to think 
 that AI could kill any language.


Some of the very worst tech ever made has as its primary feature 
that it doesn't require a programmer, and invariably the result 
is so difficult that you still need a programmer, but now your 
programmer doesn't have any of the tooling or standards or 
documentation that programmers expect. Google has a SOAR solution 
for example that has a log-parsing configuration language which 
is a Lovecraftian sanity-threatening horror to contemplate, and 
everywhere in admin IT there are enormous "configurations" with 
inscrutable mutating state.

With even the worst desktop AI, like a two-year-old model running 
in ollama, there's no longer any plausible reason for any of this 
garbage to exist. You can't tell a fable anymore about how domain 
experts can "just" edit some JSON instead of touch code, because 
AI would let anyone work with a configuration library used by a 
standard programming language, and easier and faster and more 
reliably than it can help with these NIH DSLs.

For this reason I'm mostly optimistic about AI killing languages. 
Like a fever, or a mild poison, it'll kill this stuff first.

That it makes legacy code easier to escape is a good thing for 
better languages, and it doesn't seem to have been bad enough at 
less-popular languages to provoke a Python rewrite of the dlang 
forums.
Feb 10
prev sibling parent Lance Bachmeier <no spam.net> writes:
On Tuesday, 10 February 2026 at 17:38:40 UTC, matheus wrote:


 Interesting, since I'm not using AI I'd like to know, in this 
 case you have LLM locally and you point to D source folder and 
 It learns from that database and do everything from there?


You *can* run LLMs locally, but it depends what you want it to do 
for you. I use llama with qwen3-coder-next on my desktop CPU. It 
works fine for asking it about what a function does, what some 
code does, writing short functions, or documentation of short 
functions. You no longer need a GPU for anything like that. Just 
a free model and a computer with enough RAM.

If you want to do anything that requires thinking or a lot of 
context, AI services like Claude or Gemini are still your best 
bet, or you better have an expensive computer sitting around 
doing nothing.


 they showed us a port of one module written over the past 20 
 years in a new language using AI in just a couple of hours, it 
 was modernized and everything else through AI.


If you want to port C code to D, that's to a large extent already 
done. I can convert C header files quickly using ImportC plus 
Gemini for cleaning up the edge cases. I'd be very cautious about 
"modernizing" C code with an LLM. You better write an extensive 
test suite before you try that.
Feb 10
prev sibling parent Paolo Invernizzi <paolo.invernizzi gmail.com> writes:
On Tuesday, 10 February 2026 at 16:14:03 UTC, Vladimir Panteleev 
wrote:

 On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 [...]


 I guess I could post a few thoughts about AI / LLMs here if 
 people are interested.

 [...]


I'm on hurry now, but same feeling here with Claude-code: Opus 
4.5 changed everything, and it really good in writing D code.

We've curated it's configuration a lot (for example, pointing it 
to prefer design-by-introspection and CT and so on ...) and it's 
doing great!

I think that CT is a big plus, as it can iterate fast catching 
errors directly trying to compile (that's why it works very well 
also in Elm!).

It improved a lot also when we added something like: after having 
coded a library implementation, span an agent and tell him to 
build something with the library, then look at its performance 
and improve the point that were confusing for him.

/P
Feb 11
prev sibling next sibling parent Kapendev <alexandroskapretsos gmail.com> writes:
On Monday, 9 February 2026 at 21:25:02 UTC, user1234 wrote:

 For some reasons I monitor a full hand of programming languages 
 repos.This is interesting because I can compare to others.

 One tendency I have noticed recently in the D world is one guy 
 that is very good with AI. Cybershadow. Already 5 or 6 PR, he 
 masters the tools.

 One problem I'd like to mention is that if for "this" or "that" 
 AI may help, you have the risk of not understanding the 
 codebase anymore, after merging 100 PR, let's say, the 
 knowledge will be lost or costly to re-acquire.

 The paradox is however that often human-generated PR are just 
 patches. The compiler is a huge patchwork, even if there's many 
 sub programs ( visit this or that ).

 So you see it's a bit the same. People come and go, newcomers 
 have to understand what people who left did. This is what 
 matters.


My code is 100% vibe coded, from start to finish.
It's amazing really. I can watch an episode of The Walking Dead 
on one of my 4K screens and have Gemini CLI code for me on one of 
my 1080p screens.
I think nobody noticed here that I was doing that. Well, I am 
sharing it now with you people. I am enjoying the gift that is AI 
for just $2.99!?!!
Feb 10
prev sibling parent reply Guillaume Piolat <first.name gmail.com> writes:
Personally I like using Cursor for tooling but only use them one 
month out of two, just in case.

Some non-trivial downsides with AI, from what I heard:

- it's a good feeling of going faster, however 96% of people 
don't trust the code but 48% check it, which means the tools 
encourage you not to read too much what it generates. At least 
when you write sme code, there is baseline level of caring about 
it.
- as you can go faster, you also create more stuff to be busy and 
it can be more intense
- working without it becomes feels a bit raw and "new", as if you 
forgot how to walk
- if the LLM are left to their own devices, as happens in 
autonomous agent communities, they end up doing a whole lot of 
nothing, exchanging a-ha moments and manifestos rather than 
action. They also don't tend to test what they do, since they 
don't care about reputation. LLMs are very much gullible.
Feb 11
parent monkyyy <crazymonkyyy gmail.com> writes:
On Wednesday, 11 February 2026 at 21:47:16 UTC, Guillaume Piolat 
wrote:

 - if the LLM are left to their own devices, as happens in 
 autonomous agent communities,


That was fake and a marketing psyop. Gpt 2 level of text 
processing with promts like "pretend to be an scifi ai gaining 
sentience" "pretend to be planning on killing humanity" and then 
the cyptro scam bot farms did a quick rewrite of their tools to 
use the api and it was flooded with that.

Everyone got what they wanted to see.
Feb 11