ROOL Forum: Recent posts

Sun, 04 Oct 2026 22:17:16 +0000

What’s the copyright status of the online PRMs?

Sun, 04 Oct 2026 21:40:50 +0000

With reference to https://www.riscosopen.org/forum/forums/1/topics/20578/posts/166521 - does anyone else agree that the reference to the Ack on the Wiki is incorrect? If so, I'll edit it.

Sun, 04 Oct 2026 21:34:59 +0000

@David J R

When you all have is a hammer, everything is a nail [...]

Please do be careful with the strong anti-AI advocacy as at times it can come off as a personal attack. In part, that can be down to people taking something personally but given the context and to-and-fro over the last 8 pages, it's easy to see how it could be a rather directed comment, not a general one, on your part.

For what it's worth, personally (not ROOL) I really hate reviewing most AI generated code. I hate the tone of voice in the borderline incomprehensible mangled "English" comments. I truly despite the grotesque, deeply corrupt companies involved. I hate the way it turned what little progress we might have been making on climate change into an even bigger environmental disaster. And for all that damage, I personally think the quality of the output is typically pretty poor. Any non-trivial project that runs for a few months seems to always hit the same wall of spinning on fix-one-thing-break-another and the maintainers start to notice that the cost of tokens to keep the whole thing going is quite painful... And we're well aware that the cost is far below what it should be for the big companies to break even.

Open models offer alternatives, but the local hardware requirements are absurd and it's unlikely they'll ever match performance of models operating on vast server farms. The whole LLM parlour trick is, after all, an outcome based on sheer brute force. And even if you did see a shift to on-device local models, would they - could they - be trained ethically?

And for all these things I hate? The majority of the many millions of people on Earth who have access to this technology apparently don't care. They simply don't share my views and values in this matter.

Best we can do is focus more on our local politics, vote with our ballot boxes and our feet in terms of what we choose to use, and largely ignore what everyone else is doing where possible. If you want to become involved in political movements to try and spread the message of AI shortcomings and dangers more widely, then great - but the ROOL forums are a tiny, tiny corner of the web and it just won't have any impact to repeat this stuff here. If someone else is using AI to make something and other people are happy with it, then just let them get on with it - unless licence conditions are being breached, of course.

Sun, 04 Oct 2026 21:22:47 +0000

I must be missing something obvious, but why? What use does that serve, beyond creating a PDF that can’t be edited when corrections are made to the Wiki?

It will have been created by Martin because the ROOL wiki can’t be digested by LLMs due to them being blocked, but the PDF can.

Sun, 04 Oct 2026 21:20:17 +0000

Which is a copy of the ROOL Wiki, if this is incorrect I’ll update my reference information at my end.

We did this one a couple of weeks ago... https://www.riscosopen.org/forum/forums/5/topics/20558

TL;DR: I checked the Wimp source code, and the Wimp sends the messages unrecorded so no Ack is required; no other documentation apart from the ROOL Wiki mentions the Ack; the text on the TaskWindow_Output page looks very much like a copy from TaskWindow_NewTask, which does (by its nature) require an Ack in order to work; the strong suspicion is therefore that the Ack note somehow moved between pages.

I did wonder at the time whether I should have suggested that I update the Wiki, but didn't. So what do the collective (non-AI) folk think? Should the Ack reference be removed from the Wiki, as it seems to be on its own and wrong? I'm always a little cautious about going in and changing stuff like this, as the person who wrote it may have had a very good reason for doing so that I'm not aware of.

Which is a copy of the ROOL Wiki

I must be missing something obvious, but why? What use does that serve, beyond creating a PDF that can't be edited when corrections are made to the Wiki?

Sun, 04 Oct 2026 21:19:18 +0000

@Steve P

I think the statement, from Andrew H, of no AI code in OS submissions was pretty definitive.
I did not say that at all [...]
That was my understanding of these words:

You use AI to help draft tests, type quicker than you can, assist in understanding your domain and so-on? Good for you, let’s see what you’ve got. You use AI for cognitive surrender, believing it can think, and understand, and writes good code? ROOL is not interested in your MR.
Maybe you weren’t as definitive as I thought.

"You use AI to help draft tests, type quicker than you can, assist in understanding your domain and so-on?" is pretty clear that AI is writing some of the code. It's not vibing anything, but it's still assisting - as some have said, albeit a rather tired colloquialism these days - you can use it as "autocomplete on steroids".

We've seen submissions that were fully vibed and they call into the "thousands of lines" category because the author (in those cases, quite openly) didn't understand the domain. But, of course, neither did the LLM. It produced something that seemed to work superficially, but fell apart quickly as soon as you went even slightly off the well-trodden happy path.

For ROOL to say "You can't use any AI tools whatsoever and we'll accept no submissions using such" would be daft, because it would be unenforceable - there's simply no way to detect use of AI to assist and accelerate in development when driven by someone who understands their domain and reviews, and modifies where necessary, any automatically generated output.

Sun, 04 Oct 2026 20:49:34 +0000

The Configure plugin is a big improvement.

I’m in the process of trying it with a PS4 controller.

Sun, 04 Oct 2026 20:37:35 +0000

Great thing is there are loads of super-cheap controllers available. I'd be interested in hearing feedback with different devices (I usually use a cheapo clone of an XBOX360 wireless controller).

I've actually released 0.24 now.
https://forums.jaspp.org.uk/forum/viewtopic.php?t=731

Sun, 04 Oct 2026 20:24:36 +0000

User_Message_Acknowledge “This code is returned from Wimp_Poll and indicates that the message sent from this task was not acknowledged by any recipients.”

https://www.riscosopen.org/wiki/documentation/show/Use…

Sun, 04 Oct 2026 20:09:35 +0000

The printed RISC OS 3 PRMs p. 266 say nothing about acknowledging TaskWindow_Output. Ditto online version from www.riscos.com.

It’s on page 1361 of:

https://www.riscostechnologies.co.uk/downloads/risc-os…

Which is a copy of the ROOL Wiki, if this is incorrect I’ll update my reference information at my end.

Sun, 04 Oct 2026 19:53:59 +0000

I also had the code checked against the PRMs. One real issue came up: TaskWindow output messages weren't being acknowledged as the PRM asks. That's fixed.

The printed RISC OS 3 PRMs p. 266 say nothing about acknowledging TaskWindow_Output. Ditto online version from www.riscos.com.

My non-AI powered human brain Mk1 managed to recall that yesterday, in catching up on the forum, I saw a thread from Steve Fryatt covering exactly this question. The thread was prompted by Steve's excellent non-AI powered human brain noticing a discrepancy in the online documentation.

Ironically, after Steve checked back to the RISC OS source code, which confirms that acknowledging is totally unnecessary, he said "Perhaps I should have asked AI? :-)".

I think we now can see that AI would have given him the wrong answer!