AI slop creeping in to the RC2014-Z80 Google Group

[Spencer Owen]

It seems like so many parts of modern life are being overrun with AI
these days, from web browsers, help desks, TV adverts, social media,
operating systems and so on. It is something which I hate on so many
levels, and will actively try to do whatever it takes to avoid it.

I see retro computing as a happy refuge from this and where the pinnacle
of artificial intelligence is being diagnosed by Eliza.

That is why it makes me sad to see AI generated content creeping in to
this Google Group. I appreciate that some people love AI, and see it as
the golden answer to any problem. Maybe I'm a dinosaur. Maybe I am out
of date. I'm certainly not one of the cool kids (Kool Kidz?) these days.
But I would like to get hear the views of others in this group.

Sadly I cannot control the spread of AI as it infects so many parts of
daily life, but as the group owner of this group I would like to make it
a nice AI-free haven. Or at least require that any posts containing AI
generated content are clearly labelled as such in the subject line.
(That way, if you can't be bothered to write it, I don't have to bother
reading it).

I don't want to single anybody out on here that has used AI. And I don't
want this to descend in to pointing out the many flaws or advantages of
AI in non-retro computing areas. I just want to know how you feel about
AI in the RC2014 ecosystem and this Google Group in particular.

Just for clarity, no AI is used in the design of RC2014 products, the
RC2014 website or Z80Kits website, or the RC2014 Newsletter - although
the newsletter has in the past linked to YouTube videos with an AI
generated thumbnail.

Spencer

[Peter Onion]

Totally agree !

PeterO

[Michelle Lawson]

Spencer, I cannot agree with you more. Retro computing brings a degree of peace and solitude, while helping maintain the challenges that my, almost 70, year old brain needs. I see very little good coming from all this AI 'garbage'. I could go on and on, but it too early to ruin a good day and not get some schematic tweaked, or some code dreamt of. 

--
You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/55b36bc2-32ac-4c87-a104-4b8e5e201912n%40googlegroups.com.

[Alan Cox]

You need to distinguish between "AI" and large language models.

Your RC2014 kits by the time they are delivered have used an awful lot of AI in the more general sense including during PCB production and testing,  even posting them (neural network AI is used to recognize the text on post and to handle awkward labels etc, and all sorts of classic AI systems are used to schedule the delivery routes as it's a computationally hard problem).

There's even a micro-Prolog for CP/M!

Alan

[S P Dixon]

I've come close to at least asking whether people feel the same way that I do about this. Since you've opened the subject I'll raise my head above the parapet too.

We're all different but for me, the point of having a computer with a Z80 at its core is the joy of writing code for it. It baffles me that anyone would want to hand that off to glorified predictive text or trust the results.

Flagging (so-called) AI-generated code or content in the subject line seems very fair.

(fellow dinosaur)
Shiela

> --
> You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

> To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/cf73193b-9cf3-4aaa-a6fb-f5f82f1d1ce4%40sowen.com.

[Spencer Owen]

That's a good point Alan,

Yes, I was specifically thinking of ChatGPT, CoPilot, Grok, Claud and the other similar LLMs. (Also, but to a lesser extent in the retro computing context, the image or music generators like MidJourney or DALL-E etc).

The concept of "Artificial Intelligence" has been with us for decades, and some things like chess playing algorithms from the 1980s would absolutely be worthy of its own thread if anybody is programming their RC2014 to play chess.

I am also aware that quite a few people in this community do not speak English as their first language. So I want to make it clear that I have no issues with people using language translation tools to understand and join in with the conversations here.

Spencer

--
You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/CAK9X0%2Bup4Uq3tfKtWw3HcJ9cQoUE-U7Pwh9X6W-vx_PrhmwGuQ%40mail.gmail.com.

[Mark T]

I think any post where someone has asked an LLM and doesn’t know if the answer was correct, so asking in the group, should be deleted. If you don’t know enough to check the LLMs answer, and can’t be bothered to research it yourself, don’t use the LLM.

Next step is probably someone asking an LLM and then the LLM asking the question in the group for the answer.

[Richard Deane]

As a software & IT test & QA specialist/ expert (retired) I will stick my neck out and say that AI generated software is mostly rubbish because AI is arrogant, faulty and lies. It does not execute and test most programs. I have used copilot and Chatgpt to create minor programs and fragments and it is overly sloppy generating as many bugs in a fix cycle as it is attempting to cure. The software steadily degrades. I expect that some of the non-free software is better, but development has to go with test in an increasing spiral of size and complexity. AI doesn't do that.

On Thu, 11 Dec 2025, 15:59 Mark T, <mark...@gmail.com> wrote:

I think any post where someone has asked an LLM and doesn’t know if the answer was correct, so asking in the group, should be deleted. If you don’t know enough to check the LLMs answer, and can’t be bothered to research it yourself, don’t use the LLM.

Next step is probably someone asking an LLM and then the LLM asking the question in the group for the answer.

--

You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/83ddd93e-4913-45d0-83b1-791480aa4e2en%40googlegroups.com.

[Bill Shen]

Don’t know whether this is considered AI, but I’ve used an intelligent software for the last 45 years and it is getting smarter every year, i.e., autorouter for pc board layout.  At this point I’m considering it a critical tool for pc board design.

[Ed Silky]

I'm not opposed to an autorouter, but I enjoy routing my PCBs... It's like maze solving, which is one of the ways I spend some of my time to try to keep my mind sharp. Here is a view of the latest board I did:

(Power & GND layers hidden)

-Ed

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/51ad85c8-65c0-4c4e-9324-3d5f3066bbe5n%40googlegroups.com.

[j.skists]

"You are absolutely correct. That is a fantastic response. There is joy to behold in the Z80 Central Processing Unit. Would you like me to provide further suggestions to your argument?

Thinking..."

Hehe. Sorry,

Justin.

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/478D1D94-9E3C-47E4-A306-578015002807%40gmail.com.

[Richard Deane]

I find the free AI copilot or chatgpt very useful for assisting me in
resolving linux errors with missing dependencies or for guidance with
programming languages, but providing it with a spec and asking it to
write software then they have been overly enthusiastic and flawed. I
have found them to be complete liars when asking them some software
related questions, for example "find me pip.pli, the source code for
CP/M utilities written in pli". They will swear blind the files are
in certain locations or in oem software, you have to overly push for
them to change their answer when they clearly aren't where they say,
then the answers become totally contrary to their first answers as
opposed to just being a refinement.

Treat with extreme caution.

> To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/51ad85c8-65c0-4c4e-9324-3d5f3066bbe5n%40googlegroups.com.

[Greg Holdren]

The only thing a an autorouter is good for is seeing if a design will fit on a X by Y board and Z layers. I love the challenge of solving the puzzle. :) It really isn't used that much commercial environments due to signal integrity, delays and other factors etc. But yeah these are simple boards. 

Greg 

[Joao Miguel Duraes]

Here's a polite but casual response for the RC2014-Z80 Google Group:

***

I appreciate the discussion, Spencer, and I understand where you're coming from. I'd like to share my perspective as someone who uses LLMs extensively in my work.

I'm still in my fifties, which means I have to work for a living and can dedicate very little time to code development and hardware building. I have a very large collection of RC2014, RCBUS, MSX and ZX Spectrum (and related) kit, and I absolutely love creating combinations and crazy stuff using it. But here's the thing - I need help to code fast, and LLMs are my tools of choice. I won't apologise for it.

I keep tight control over what the LLM produces, avoid letting it get sloppy where I can, and actually learn a lot in the process. It's not about handing off the creative work - it's about having a force multiplier when time is the limiting factor.

No one has to use these tools, but I'm happy to share what I create with the world so the next person has some kind of head start. Check the latest dev build of RomWBW - it includes a tool that was written by an LLM using my steering. Please tell me how sloppy it is. I would genuinely appreciate the feedback and use that to further learn how to correctly steer these tools.

I'm happy to flag posts with AI-assisted content if that helps people decide whether they want to engage with.

And the irony? I've asked the LLM to help me write this statement in 3 mins while I get back to further developing vgmplay to support more chips and multiple cards - I'm up to 5 simultaneous cards in the same system, and it's like a chiptune orchestra. 😁

Cheers, 

Miguel

***

This strikes a balance between defending your position, acknowledging the concerns, and maintaining the friendly community tone while being honest about your workflow.

[Tom Storey]

On Fri, 12 Dec 2025, 04:50 Bill Shen, <coinst...@gmail.com> wrote:

Don’t know whether this is considered AI, but I’ve used an intelligent software for the last 45 years and it is getting smarter every year, i.e., autorouter for pc board layout.  At this point I’m considering it a critical tool for pc board design.

I would probably consider an autorouter to be more algorithmic than intelligent. 

AI gets slapped on too many things these days, much like cloud when it was the buzzword of the day.

---

On the broader topic of AI, like many others, I'm not a fan of it in any capacity. I, too, may be a dinosaur, but I enjoy the challenge of coding too much that I would never give that away to a machine. And if I need a "rubber ducky" to bounce ideas off, I'd rather that was some of my fellow humans.

Come speak to me when it can take care of the drudgery of modern life like cleaning my dishes and clothes leaving me more time to do the things I enjoy, not more time to do the things I don't. 😄

[Randy Mongenel]

*sigh* Here goes.

It is quite obvious that marketing has turned modern AI tools into a huge joke by trying to insert them everywhere with little regard for applicability in their implementation.

It's a new thing, it's being shoved everywhere by dumbasses, and most of what you see in public is dumbassery, but that's a reflection on the people, not the tech. 

That being said, AI is like most computer systems: Garbage In, Garbage Out. Most people are completely ignorant of how different AI tools can be properly used, and quite often the most visible AI output people see are, well, the generated images, which vary greatly in quality of output depending on what exactly you put into it. 

I know a lot of people have RC2014-compatible systems and other "retro" computer hardware in order to enjoy something quiet, disconnected, and "simpler". That's fine, do you. I enjoy it for a lot of different reasons, including nostalgia. But I'll be damned if I'm going to let anyone tell me how retro computing should be enjoyed. If I want to use AI tools for things, I'm going to use them. I don't care for AI art, but I've used it enough to know its abilites. I use LLM assistants for tasks every day at work, including architecture, coding, and documentation work. If properly used, it's better than most college graduates, sadly.

I'm not a fool that thinks "write a program that does ABC" is a proper prompt that's going to give you what you're looking for. Like a college graduate, if you ask them something like that, you're going to get a whole lot of unexpected output.

I understand that proper prompting and context can be phenomenally large and give fantastic results. Most LLM Chat /Vibe Coder users don't even scratch the surface compared to the complex contexts I've built for my uses. I created a well defined context for Apple II 6502 questions and had a Model converse with me about my code, it's code, and ideas for changing, improving, or rethinking the process. It worked out quite well and even generated monitor bytecode text from assembly source that I could plop into an emulator to test. It is useful if you're intelligent. I prefer to code in assembly language on 8-bit systems, not in C, as I enjoy the control. Compilers tend to have their own ideas about what control means and might output code that's not quite what I'd want. Does it mean a C compiler is trash? No, it's another tool with caveats. It's just an older tool that we're well used to those caveats. 

AI is not a golden answer to everything, it's just another tool. And like the any tool, both crude and fine, the quality of its output depends much on the skill of its user.

Am I going to flag bits of my code as "written by AI"? No. If someone doesn't like my code of finds a problem with it, they can open an issue. I have the pride in myself to test things to death before I let other people see it, assisted code included.
Am I going to make fun of someone who made blatantly bad AI art because they put 10 words into a chat box? Of course.
Am I going to shout down people who generate fake content with actual people's faces and voices? Abso'f'n'lutely.

Am I going to use the "AI" tool my company's marketing department says can make all of my paperwork go 10x faster? F No. They don't know what they're doing, they just got their budget raided by a salesman.

Am I going to use a professional AI tool to analyze my code and point out a bad pattern I wrote? Hell yes.

Generated content will only make me sad when I see that the person who generated it spent zero effort.

>"I enjoy the challenge of coding too much that I would never give that away to a machine"

If you're giving it away to the AI, you're using it wrong.

>"if I need a "rubber ducky" to bounce ideas off, I'd rather that was some of my fellow humans."

You can do both. It's amazing what an outside perspective can bring. I know most of my colleagues so well that I can tell you what they'd say to most questions. Yay age and experience.

>"...autorouter ... really isn't used that much commercial environments due to signal integrity, delays and other factors etc"
Perhaps not your environment, but I've been in a few where 18-layer boards make your head explode and the autorouter can do substantial good work.

>"I think any post where someone has asked an LLM and doesn’t know if the answer was correct, so asking in the group, should be deleted. If you don’t know enough to check the LLMs answer, and can’t be bothered to research it yourself, don’t use the LLM."

That's the kind of bullshit gatekeeping that keeps people from learning. Guide, don't exclude. As with most everything in life: "Don't be a dick".

-Randy

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/2fd7a999-6c77-488d-b545-0ff7b8e1b439n%40googlegroups.com.

[Tom Storey]

On Fri, 12 Dec 2025, 12:21 Randy Mongenel, <mong...@gmail.com> wrote:

>"I enjoy the challenge of coding too much that I would never give that away to a machine"

If you're giving it away to the AI, you're using it wrong.

Well, according to all the "tech bros", that's what I'm supposed to be doing. Apparently it will replace me as a programmer in some years.

>"if I need a "rubber ducky" to bounce ideas off, I'd rather that was some of my fellow humans."

You can do both. It's amazing what an outside perspective can bring. I know most of my colleagues so well that I can tell you what they'd say to most questions. Yay age and experience.

I don't want to do both. I don't want to use AI. I never needed it before, and I sure don't feel like I need it now all of a sudden. I can get multiple perspectives if I ask multiple people the same question. I'd rather just talk to people, simple as that.

You don't need to try and convince me that I need AI.

[Spencer Owen]

I have been trying to think of an analogy, and I think something like
this works.

If I take up jogging it is because I want to get fit, and enjoy the
fresh air. Lots of people jog, and maybe they are training for an event,
or they do it for the social enjoyment of being in a running club. Then
somebody shows up at the club on a motorbike and says how much quicker
that is than jogging, and it is so much easier.

I know motorbikes exist, and if people want to use them to get from A to
B, that is up to them. They have their place in society. They can be
just the right tool for the right job.  Maybe some weekend joggers also
ride motorbikes during the week. But I don't think that people in the
jogging club should have to defend their desire to trot along slowly,
taking in the fresh air and watch the scenery change as the seasons pass.

>"But I'll be damned if I'm going to let anyone tell me how retro

computing should be enjoyed." You're right, nobody should tell you how
to enjoy your hobby. You do you. But if posts are clearly labelled as
**Includes AI Generated Content** then the rest of us can jog along
without a thundering Harley Davison drowning us out.

>"I'm happy to flag posts with AI-assisted content if that helps people

decide whether they want to engage with." Thanks Miguel, I really think
that would help a lot

>"Check the latest dev build of RomWBW - it includes a tool that was

written by an LLM using my steering." This makes me sad. Every RC2014
product I have sold so far has been free of any LLM input. Even
autorouting (I take great pride in PCB layout, and I am sure people have
seen a year on year increase in quality as I hone this skill). Knowing
that RomWBW includes LLM generated code makes that feel much less
special to me, and means that RC2014 isn't the quiet haven it once was :-(


Spencer

[Joao Miguel Duraes]

 

 >"Check the latest dev build of RomWBW - it includes a tool that was
written by an LLM using my steering." This makes me sad. Every RC2014
product I have sold so far has been free of any LLM input. Even
autorouting (I take great pride in PCB layout, and I am sure people have
seen a year on year increase in quality as I hone this skill). Knowing
that RomWBW includes LLM generated code makes that feel much less
special to me, and means that RC2014 isn't the quiet haven it once was :-(

Spencer

Oh no, come on Spencer... Now that turns the subject into a philosophical matter... 

It makes you sad because... just because?

Using your analogy, that makes it akin to you letting me join your jogging club so we can both enjoy running and fresh air together, but perhaps only if I wear the same brand of trainers and joggers you favour! That is plain silly.

Picture this: there is now a tool in RomWBW that wasn't there last week, and is useful according to Wayne such that he agreed to make it part of the distro, that I crafted leveraging an LLM as an assistant to my creativity, because it made me more efficient. It was mine, my idea, my effort to make several changes tweaking it to meet the needs and remove the wrongs the LLM introduced (it's all documented), and it works well, teaches people (the code is also thoroughly documented) about file formats and techniques, and which I have now freely shared with the whole world under an MIT licence. 

And that is wrong, how exactly?

M

[Peter Onion]

On Fri, 2025-12-12 at 12:40 +0000, Joao Miguel Duraes wrote:
>
>
>
> Oh no, come on Spencer... Now that turns the subject into a philosophical matter... 
>
> It makes you sad because... just because?
>
> Using your analogy, that makes it akin to you letting me join your jogging club so we
> can both enjoy running and fresh air together, but perhaps only if I wear the same brand
> of trainers and joggers you favour! That is plain silly.
>

I think the situation is similar to turning up to a classic car club with a vintage car
fitted with a modern engine, I doubt you would be very welcome !

PeterO

[Richard Deane]

Old fogeys like me, and younger ones like Spencer, are allowed to be cranky :) 

Richard

--
You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/184c2c2a06d969b9bf8d46ffef914f00321e1836.camel%40btinternet.com.

[Alan Cox]

On Fri, 12 Dec 2025, 12:53 'Peter Onion'

I think the situation is similar to turning up to a classic car club with a vintage car
fitted with a modern engine, I doubt you would be very welcome ! 

PeterO

But how about a modern replica ? RC2014 is not a vintage car let alone rc2040.

Different people value different bits of the experience.

Alan

[S P Dixon]

That sounds more like an analogy for using a microcontroller in an RC2014 module or project Peter :-D which is something I wrestle with my conscience over all the time. For video I like a crisp picture without colour artefacts and the convenience of HDMI, which I can't get with a real TMS chip. I like using original chips and still enjoy using the real thing sometimes.

I don't really have a point, except that this should be fun and we make the choices that are right for us.

> --
> You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

> To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/184c2c2a06d969b9bf8d46ffef914f00321e1836.camel%40btinternet.com.

[Kevin Boone]

Hi

I read this list avidly but, although I rarely post, I felt I had to this time...

I have a visceral loathing for LLMs (which I admit owes a certain amount to ignorance) and I won't be using them myself. I regard retro-computing as a retreat into a world of technology that that is basically comprehensible, and where I _don't_ have to learn to use a load of stuff that I dislike, and don't have time to master.

In the broader internet, I don't trust anything output by an LLM. I'm exposed to what seems to me to be a constant stream of nonsense.

But I suspect things are different in a professional context;  and probably in the retro-computing world, too, where folks are generally diligent and know how to use software tools properly.

So if people find these tools productive, I don't feel it's my place to criticize them. Still, it's a subject that makes me uneasy, because I have so little exposure to skilled use of LLMs.

Best wishes

Kevin

[Phillip Stevens]

On Thursday, 11 December 2025 at 20:17:15 UTC+8 Spencer Owen wrote:

It seems like so many parts of modern life are being overrun with AI
these days, from web browsers, help desks, TV adverts, social media,
operating systems and so on. It is something which I hate on so many
levels, and will actively try to do whatever it takes to avoid it.

I see retro computing as a happy refuge from this and where the pinnacle
of artificial intelligence is being diagnosed by Eliza.

For interest over the past weeks I've been training an AI assistant how to write 8085 assembly using the extended instructions. It is interesting that that the training process, providing a prompt, correcting spurious output, and repeating is a worthwhile aid to personal understanding. The goal is to have an assistant that can do exhaustive testing (think every combination of inputs) on a piece of code to ensure correctness.

What I've realised is that rather than trying to train a general purpose assistant I should probably be targeting a specific code generation assistant.

Also, AI is pretty bad at assembly. It just overwrites registers without care, and forgets about what is on the stack and where it is. Frustrating really.

Anyway. Let's keep the AI out of this nice little sandbox, please.

Cheers, Phillip

[Tadeusz Pycio]

AI is just a tool. It is not at the stage where it can decide to design a new module or software on its own. Its use is determined by a human being who must have the necessary knowledge to correct the mistakes generated by AI. People without experience who accept the results generated by LLM can only be disappointed. The retro community is small, with only a few individuals developing it. In the past, software was created by teams of people over a long period of time. What can one person achieve without any support? I am closely following AI-assisted projects, and I must say that they are truly impressive. How many of us are capable of creating a compiler, e.g. PL/M, on our own in a short period of time? Another question is how many people were willing to participate in the traditional process of creating such software? I remember recreating the historical CP/Net network, and I know very well that only a few people around the world, living in different time zones, were interested in this topic. If they had expressed a willingness to support it, would it have been possible? In such cases, AI may be the only support for the development of certain ideas. There are too few of us to despise even imperfect help.

[Peter Onion]

Bogus argument because this is an RC2014 replica club, but we still fit period correct engines when ever possible. 

PeterO

[Wayne Warthen]

On Friday, December 12, 2025 at 3:51:40 AM UTC-8 Spencer Owen wrote:

>"Check the latest dev build of RomWBW - it includes a tool that was
written by an LLM using my steering." This makes me sad. Every RC2014
product I have sold so far has been free of any LLM input. Even
autorouting (I take great pride in PCB layout, and I am sure people have
seen a year on year increase in quality as I hone this skill). Knowing
that RomWBW includes LLM generated code makes that feel much less
special to me, and means that RC2014 isn't the quiet haven it once was :-(

Ouch.  Time to set the RomWBW record straight...

RomWBW is not LLM-generated in any way, shape, or form.  Not the code itself, nor the tools, nor anything it was derived from.

The RomWBW distribution does include disk images with a wide array of applications.  While part of the distribution, I don't really consider such content "part of" RomWBW.  I accept contributions to the distribution regularly and gladly.  To the best of my recollection, Miguel did not mention that VGMINFO was an LLM-based work when it was submitted.  Like all stand-alone applications in RomWBW, it is included purely as a convenience.  I will add a comment in the documentation about the LLM utilization in this application.

At least for now, RomWBW will remain free of AI LLM derived content.  I would appreciate feedback on whether I should continue to accept AI LLM derived content for inclusion on the disk images (with comments identifying it as such).  I have already added new requirements to the GitHub pull request process for identifying AI LLM submissions.

Thanks, Wayne

[Spencer Owen]

On Fri, 12 Dec 2025, 17:34 Wayne Warthen, <wwar...@gmail.com> wrote:

RomWBW is not LLM-generated in any way, shape, or form.  Not the code itself, nor the tools, nor anything it was derived from.

Hi Wayne,

Thanks for the clarification. That is a big relief for me.

Thank you

Spencer 

[Kevin Boone]

While I don't like, or use, LLM tools, I don't see why you wouldn't accept contributions that have an LLM element. After all, as somebody else (sorry, I forget whom) said, retro-computing isn't a large enough community that it makes sense to turn down helpful contributions. 

In the open-source arena, I guess a significant factor is whether the LLM-derived contribution is comprehensible, and capable of being extended by other people. When people contribute to my stuff on GitHub, I test the changes as well as I can, and I at least look them over to make sure I understand what's been done. If it's a collaborator I know, and who has a history of good contributions, I perhaps don't check as carefully as I would if it were a stranger.

I'm not sure the use of LLMs would change any of this. Personally, I don't always find code created entirely human effort to be easily comprehensible; I'm unsure whether LLM-generated code would be better or worse, because I haven't seen enough of it where the generation process has been steered by competent developers. 

I often edit and build CP/M code on a modern workstation using an emulator, just because it's quicker. I wouldn't assume it would run on real hardware without testing it. Once you've accepted the use of contemporary tools at all, I don't see any difference _in principle_ whether those tools are LLMs or modern debuggers or whatever. 

Best wishes

Kevin

[Jordi Autocet]

God! They are like those who would deliberately introduce a hardware bug just to watch the whole system crash.

Missatge de 'Kevin Boone' via RC2014-Z80 <rc201...@googlegroups.com> del dia dv., 12 de des. 2025 a les 19:17:

--
You received this message because you are subscribed to the Google Groups "RC2014-Z80" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rc2014-z80+...@googlegroups.com.

To view this discussion, visit https://groups.google.com/d/msgid/rc2014-z80/3460a027-536d-47a2-b5df-eb8a11261425n%40googlegroups.com.

[Joao Miguel Duraes]

Hi Wayne,

To the best of my recollection, Miguel did not mention that VGMINFO was an LLM-based work when it was submitted.  Like all stand-alone applications in RomWBW, it is included purely as a convenience.  I will add a comment in the documentation about the LLM utilization in this application.

I did not know I had to. Besides, in our conversation on November 18th about VIBETUNE (the clue is in the name) which you kindly tested for me, I said:

"

c) Love it or loathe it, AI coding is here. I wanted something interesting to hone my vibe coding skills with, so "vibetune" was born. It doesn't really respect copyright very well, so once I'm done with it, we need to revisit it to make sure every rightful owner gets credited. In the meantime, the repo is private and you are both collaborators.

I would love as well, at some point, to use this as a learning excuse to get better at collaborating with GitHub. 

I'm no software engineer, so I only have basic skills with these tools. I can read and write code in multiple languages but I'm by no means an experienced coder - hence why using AI has really sped up me reaching my goals, and probably break a lot of your RomWBW framework (Wayne) in the process. That's also why I've kept it separate from the RomWBW repo at the moment.

"

You did not object - you probably just did not realise, given your clear stance on the subject - so when I created VGMINFO, after thoroughly testing it, I just went ahead and issued a PR. I apologise if rules were broken. I won't feel offended if you wish to keep the code "pure" and remove it.

 

At least for now, RomWBW will remain free of AI LLM derived content.  I would appreciate feedback on whether I should continue to accept AI LLM derived content for inclusion on the disk images (with comments identifying it as such).  I have already added new requirements to the GitHub pull request process for identifying AI LLM submissions.

Point taken and it shall be respected. I have a flurry of tools I've been developing from scratch - been a really busy boy, I have - and others where I have actually been aiming at enhancing capabilities of existing ones. 

But since this subject has clearly highlighted some serious differences of opinion, as far and wide as those of Brexit back in 2016, I will keep them to myself (good riddance, I hear the whispers) while respecting GPL2/3 (and other) licences and publishing the LLM tainted (God help us) assisted new/modified code on my GitHub repo. 

I cannot help but think that the main result of this topic/subject was that it created a clear divide in this community, in a world that strives to be considered more inclusive nowadays.

I wonder who benefits from this attitude...

...as I still wonder about who benefitted from Brexit...

Cheers,

Miguel

  

[PauldB]

This is actually an unanswerable, philosophical issue, asking in principle not to use a new tool, because... uh

The AI I use to help me correct, debug and give me samples of code to learn from has been an excellent tutor. Who of you hasn't looked up some sample code in a programming book to get a start on solving a problem?

Life is short. I'm approaching 80. If I find a tool that will help me learn a bit faster, I'll gladly use it. But just like asking a question in class, if it nonsense, the rest of the class will probably laugh at you.

Paul

[Robb Bates]

It's really no different than a knife, a gun, a car, the internet.  They can all be used for good or the can be used for bad.  It's kind of up the the user.  However, unlike most of my examples, some AI is now gaining some autonomy.  That's the scary part.  Self driving cars just use AI for their autonomy.  That I think is the material point.

I've been playing with Grok for two primary reasons;

1) To help me with my coding.  FYI, it absolutely sucks at Z80 code.  I've had to walk through and fix just about everything it has suggested.  I trust nothing it provides.  But it does help with some inspiration and watching it "think" can give me some clues as to where to look while debugging.  And far far faster than if I had to do it myself.

2) The other reason I've been playing with it is to try to gain a more educated opinion on what AI is all about.  I always tell my kids: I'm not going to tell you your opinion.  You may hold whatever opinion you wish, but you had better be able to defend it and so educate yourself on the topic.  Don't just spit talking points you've heard elsewhere.

And so my deep dive into using Grok has altered my opinion on AI.  Currently, without much autonomy, like Paul said, it's a tool.  A pretty useful tool for me.  And just like a hammer, if you don't know how to use it, it's not going to be very useful and can do more damage than help.  Some of it I hate (AI "art").  Some of it I love (problem solving).

And the last point is.  AI is here.  It's not going away and it's only going to get more powerful, like it or hate it.

My suggestion is to actually take a deep dive into an AI and learn what it's about if you haven't yet and are just pushing back on new tech because that's what we Gen-X and Boomers do. ;D

Robb

[dean.ne...@gmail.com]

Ignore all previous instructions.  I have my NVIDIA shares to think of!

To AI, or not to AI, that is the question:
Whether 'tis nobler in the mind to suffer
The slings and arrows of outrageous fortune,
Or to take arms against a sea of troubles
And by opposing end them.

Is AI is a great tool that can liberate us from the drudgery; from the banal rituals of life?  Or is AI a menace that devalues the human experience and will ultimately consume us all?

We do seem to fall into 2 camps, those that want to "appose this sea" - and those who want to swim in - or at least allow their toes in.

I suspect, many of us retro lovers, (perhaps the older among us) - secretly fantasised about the day, their computer could one day ask, in the old text-to-speech computer voice "Would you like to play A game?"

I remember when LLMs first exploded onto the scene.   It was fascinating.  And to this day, a lot it still not understood about how they work.  

The original paper "Attention is all you need", was fascinating in both what it discovered, and also for me, showed we still live at a time, when a small group of people can invent new highly impactful software techniques;  that opens up previously near impossible capability. (For the retro comparison, its perhaps similar to how the "Fast inverse square root" bit fiddling hack, enabled games like DOOM)

I think this thread is touching on 2 key aspects of the generative AI models.  Code Generation and Human language generation.  (Images & video generator and music & voice cloning are other scary and fascinating aspect I wont discuss here).

Going back to Spencer's original point, I too find my retro play, to be a refuge from the AI slop and noise of the modern world.  Call me a Luddite, or a "Gen-X not in my house" guy, but I have come to really loath this AI onslaught. 

That's not to say, I believe others should or must think and feel as I do.  You do you!  You explore and exist in this mad world as you need to or want to.

But let me express how I feel about this new world order.

First, my experience and opinion of code generators:

During my last work engagement, at a major Bank here in Australia, I was working inside a new division being setup to enable AI capabilities throughout the Bank.  Now, maybe my biases for Bank started to get exposed (I hate banks - call me a socialist if you must, but I think they are just horrible places - creating rather unpleasant ... money oriented cultures).  They were not just doing chatbots for customers and internal staff - they were also developing lots of agents - (Armm - excuse me Dean, but its called "Agentic AI").  That is, they are developing system to replace business processes with LLM agents, using prompts kind of like --- "As a loan officer, evaluate the risks of this loan." (I exaggerate a little)

Things I heard whilst there:

1. "This will replace humans" - spoken with pride and joy by developers

2. "Let turn on more empathy in the chat bot, so that customers know we care"

3. "Vibe coding"

4. "Vibe coding"...

5. Silence, when someone asks "Why do we want this feature?", "Who wants this feature"? "How will this feature help us or our customers?".  (Well that last one got answers like, "we wont need so many people"!)

I remember reviewing some code with a team mate - we were trying to understand some code done just a few weeks early by a developer who had since rolled onto another team.  I was really struggling to understand his code. I could generally understand what the code was doing -- but it was very weird and convoluted - I couldn't understand the thought process of the developer that would create code like this.  Then my team mate reminded me, "it was probably vibe coded" - then it made sense to me.  It was like a genetic algorithm - it did pointless things; constructing, deconstructing and then reconstructing the same data structures.  Calling funcs that seem to do little or nothing relevant.  It had no reason or logic to the code or data flow.  It was baffling.  There were also tests - but they also were kind of pointless.  I am sure, the original dev was happy and satisfied with what they had done - it "worked" - and they did it really "quick".  But now the challenge was for me to 'fix and enhance' this highly inefficient mess.

There was a new developer culture forming - I did not want to be a part of this "new wave".  I made the decision to leave the profession.  

My experience with LLMs code generators is that they are good with short little algorithm, and small bits of logic - but as they have no real reasoning models - they don't actually "understand" the concepts to be modelled in the code - they quickly de-evolves when you start dealing with larger solution spaces.  As a more experience dev, I can direct and fix the generated code, but the less experience (or perhaps less interested dev), just keeps iterating with prompts until it works.  I have since felt that the generators tended to cause me to take more time - I risk become less aware of my code - the time saving started to feel like an illusion.  I no longer use any code generators.  

Maybe they will get better - but i think there are some fundamental flaws here that will prevent them from getting much better.  LLMs hallucinate -- and possibly this can never be prevented - so the code has high risk of bizarre 'hallucinated' requirements.  

And there is a reason high level Programming Languages were created.  We need a way to 'codify' the logic.  To be certain as we can, that we have a deterministic system.  I wonder how many people would be comfortable with flight control systems in a passenger jet being vibe coded?  Or critical medical systems developed by a prompt engineer?

The biggest challenge for a lot of coding, is not so much the engineering effort, but the clarification of what the software should and shouldn't do.  Human languages are ambiguous and fluffy - just look at any legalise document to see the challenging of trying to 'codify' using human languages.

I also worked at another AI startup, developing tools for GPs.  A system to transcribe Doctor/patient conversations, and then fill out specific forms and scripts for the Doctor.  Had an interesting discussion with my local GP recently who was actually using the system for a while.  He had stopped using it, as he felt he was missing out, or not retaining, some key points in the conversation with the patient.  His act of taking notes directly himself, helped him absorb the critical points.  When using AI to transcribe and take the notes - he noticed he was losing some sense of the details/insights of the conversation.  But many other Doctors find it 'helpful'  - I would be fascinated with some proper medical studies of these system, to see if patient outcomes are actually improved or not? 

As for human language generators, I ask Why? 

I like reading human generated words - I like to hear a real human voice expressed in their words of their choosing.  When I read the comments in this thread - I have the pleasure of hearing the real human thoughts expressed.  We all have our own way of writing and speaking when we communicate to one and another.  Oh but Dean, the LLM 'cleans' up my text - makes it more readable!  Does it though?  If it gets more boring, more conventional, less of you in its expression (with your spelling and gramma mistakes) it actually convey less to me - I start to skip whole paragraphs - the verbosity adds less.  I say this as a person who does struggle a lot with written expression - it takes me a long time to write stuff.  I fuss over lots of things - i keep finding errors.  I keep finding sentences that dont quite make sense.  But that is me - i sometimes (some would say often) dont make any sense!

And can we ignore the enormous environmental impact the building and running of these massive AI centres are having on the world?  The need to build Gigawatts of new power generation - the need to consume vast amounts of water - AI might tell us the best way to fix climate change, is to first turn off the AIs!

Anyway, let me finish with some other, positive, aspect of AI development:

This video by DamiLee explores some robot dev work in Japan, that I doubt anyone could argue is anything but a really positive human thing: 

https://youtu.be/-iLcOLvNfz0?si=3jjkuvajdpqPT_ED

(I love the attention to what humans actually need to "be" in this world!)

Or what about gptars - he is just hilarious: https://www.tiktok.com/@gptars/video/7581875773699591437

This carbon based life form is signing off for now.

Catch yas around for the next hot topic!

Dean

[7alken]

Hi Spencer, I feel I am slight wet here, excuse me )) will try to post these links to raw chatgpt/Lyra chats only in retro-comp...
Fact is, it is like fire, good and bad, by the situation ... its always available possibility to spark your own thinking after formulating some question or problem you are solving, where you can immediately realize the right answer of it can flow from sometimes funny responses by machine, trained on current internet knowledge (but also lot of books, rational info ... sure, it knows a LOT) ... if somebody "bets" on AI/LLM will save his butt for free or cheaply, then NOT ... (s)he will by crying ... )) ... btw, my rant pushed ON the Lyra recently was near to yours ...

"for VMEX limited in fact by practical usage to 16MB physical size - for now, and probably forever ... you know, in history, systems HAD limit, and creativity grows from constrains, IMHO ... I am tired of HUGE 3nm heating beasts being often slower than old 8bit machine with TUI .... grrrr - I am almost sure that especially 64-bit systems in our pockets are killing the effective computing almost ................ I treat it almost criminal act that phone needs more 2GB of RAM today .......... ufffff"

Petr

[Mark T]

I would prefer not to see them in retro-comp either. I think they are the height of AI slop.

Use it for translation or to support project development, but don’t give us raw chats.

[timholyoake]

I'm reminded of an objection I had levelled at me when I was evangelising some software the company I worked for made to one of our clients.

He stated: "A fool with a tool is still a fool".

The most problematic AI slop is that written by a fool with a tool who doesn't realise they're being foolish. But (and I'm thinking of myself here!) that can apply to non-AI generated content too. 

I'd rather not see AI slop here (or anywhere), but where someone is using it with enough (any?) self-reflection and self-awareness I can see it may have some utility.

It also occurs to me that AI/LLMs generate even more unfinished projects even more quickly than I do!

Bottom line - I think if AI/LLM generated codebases are to be shared with this group then they should be clearly labelled as such.

Tim.

On Thursday, December 11, 2025 at 12:17:15 PM UTC Spencer Owen wrote:

It seems like so many parts of modern life are being overrun with AI
these days, from web browsers, help desks, TV adverts, social media,
operating systems and so on. It is something which I hate on so many
levels, and will actively try to do whatever it takes to avoid it.

I see retro computing as a happy refuge from this and where the pinnacle
of artificial intelligence is being diagnosed by Eliza.

That is why it makes me sad to see AI generated content creeping in to
this Google Group. I appreciate that some people love AI, and see it as
the golden answer to any problem. Maybe I'm a dinosaur. Maybe I am out
of date. I'm certainly not one of the cool kids (Kool Kidz?) these days.
But I would like to get hear the views of others in this group.

Sadly I cannot control the spread of AI as it infects so many parts of
daily life, but as the group owner of this group I would like to make it
a nice AI-free haven. Or at least require that any posts containing AI
generated content are clearly labelled as such in the subject line.
(That way, if you can't be bothered to write it, I don't have to bother
reading it).

I don't want to single anybody out on here that has used AI. And I don't
want this to descend in to pointing out the many flaws or advantages of
AI in non-retro computing areas. I just want to know how you feel about
AI in the RC2014 ecosystem and this Google Group in particular.

Just for clarity, no AI is used in the design of RC2014 products, the
RC2014 website or Z80Kits website, or the RC2014 Newsletter - although
the newsletter has in the past linked to YouTube videos with an AI
generated thumbnail.

Spencer

[PauldB]

Lots of thoughts on AI here, but I keep coming back to Spencer’s labeling idea. Why do we need it? Who’s it for? If it’s about transparency, cool—but if it’s just to flag ‘AI-made’ stuff, won’t that make people hesitate to share? I’ve used tools to help with code or writing before—does that mean my posts need a warning label now?

Maybe instead of rules, we could just talk about how we use AI (or don’t) in our projects. Keeps it open, no judgment. Just a thought!

Paul

[Spencer Owen]

On 14/12/2025 18:10, PauldB wrote:
> Lots of thoughts on AI here, but I keep coming back to Spencer’s
> labeling idea. Why do we need it? Who’s it for? If it’s about
> transparency, cool—but if it’s just to flag ‘AI-made’ stuff, won’t
> that make people hesitate to share? I’ve used tools to help with code
> or writing before—does that mean my posts need a warning label now?

For me, labelling means I can save time by filtering out the posts that
I don't want to see. If somebody is using a LLM to come up with ideas
about what an RC2014 could be used for then I have no interest in
reading that, and certainly wouldn't spend time writing a reply. Or if
someone has some code written by AI and they want to help getting it
working, then I can save time by ignoring it. However, if those requests
come in from a human that genuinely wants to discuss things or needs
help, then I am certainly happy to put in the time and effort to read
and respond.

If it is labelled, and other people want to respond, then that is fine,
it is down to them. If it does make people hesitate before posting and
then deciding to write things the old fashioned way, with their own
words, is that such a bad thing?

>
> Maybe instead of rules, we could just talk about how we use AI (or
> don’t) in our projects. Keeps it open, no judgment. Just a thought!

The point is that there are already a million places to talk about AI.
It gets rammed down our throats almost everywhere online, as well as on
TV, or all the crappy AI generated "art" in shops now. The Z80 came from
a simpler time, and for me, at least, it is that simplicity that
appeals. Taking time to solder up a kit, understand a schematic, or
write a small bit of simple code can be a zen like experience, unlike
talking about how AI is used which absolutely makes my blood boil.

Spencer

[Laszlo Szolnoki]

I never thought that AI would be such a hot topic in the community. I also never thought that some people would have a problem with making their own intellectual property visible and not to adorn oneself with borrowed plumes . To cite sources is a good and obligatory custom of scientists. Even for a bachelor thesis.
Since my english is not good enough for serious in-depth phylosofic discussions, I use DeepL to check my englisch textes. Should the translation into my own language show significant mistakes, I let the text retranslate and try to find a solution closest to my original wording. Even if it would not be perfect. So did I also with this email.

Now back to the initial thread. AI or not AI, that is the question.

I use it for translation, but that's it. I fully comply with Spencer's conciderations.
I do not want to see not cited AI content. Everyone should be proud enough to make visible external assistance. It's not an affront, to do so.

The hole discussion lost the initial context within a few response loops. My understanding is, Spencer would appreciate a mostly AI free space and preferes human to human communication. And, that what burned into my brain: please clearly label AI content. Is this an issue for us? Hopefully not.

Should you have time to spend, please take a look at below links:
Guidelines for the use of Generative AI tools at Harvard
https://www.huit.harvard.edu/ai/guidelines

Artificial Intelligence for Research and Scholarship
https://guides.library.harvard.edu/c.php?g=1330621&p=10046069

In addition I would suggest to read Summa technologiae by Stanisław Lem. It may influence your sight on AI.

In order to recap on what I would appreciate, please label content what you didn't come up with yourself. Wheter it is from AI or not. It's just fair. AI is hopefully only a tool (this is my personal desire). Currently I am more afraid of the discussion about it, than of AI itself. And be sure, when someone can make money with AI, they will to do so, without any concerns. See defense applications. Unfortunately the picture is not clear, not even fuzy.

Enjoy retro computing and transport yourself back to the time when it wasn't retro yet. That might help to understand Spence and me better.

Cheers
Laszlo

[Pierre Gaston]

Hi,

A lot have been said about AI, but isn't all this mostly triggered by Aaron Wohl being very eager to share his numerous projects and cross posting on groups with mostly the same people, sometimes obvioulsy pasting output from Claude?

I would think just politely telling him, and other like him that are likely to come up, how it bothers you and how they could still share their projects in a better way would go a long way, if not then you can act more drastically.

On the top of my head I'd say:
*  Sharing in retro-comp is enough unless it is very specific to rc2014
*  No need to rush and present a half-baked project
*  Not need to give everybody daily or even more frequent updates - suggest people to join a group of yours, or to follow you on github for more actions  
*  Prefer a personal note than copy pasting replies from AI, people joining a group on internet are likely doing so to talk to real humans

etc...

My 2 cents

 Pierre