Announcing QUniLator

[Hans Hübner]

Hello,

over the last couple of weeks, we have built a new version of the QBone/UniBone software.  It has deviated enough from the original software to deserve a new name:  QUniLator.  The most visible change is that it has a web-based interface which makes most of the functionality of the command line interface available in an easily browsable UI.  Some highlights:

 

 - Improved image handling for disk emualtions, including copy-on-write functionality

 - Debug workbench with memory debugger and disassembler

 - Automatically built SD card images and debian packages based on Debian 13

 - User-friendly SD card installer with network and Linux user configuration

 - New device emulations (TS11/TSV05, TQK50/TMSCP, KW11-P, MXV11-B2, DEUNA, VCB01/QVSS; DZV11, DHV11 and the RL/RLV12), validated against XXDP diagnostics where appliicable

 - New processor emulations: KA780 (VAX-11/780) and KD11-EA (PDP-11/34)

 - Several bug fixes

We have been using Claude Code to implement everything.

The new software is available on GitHub at https://github.com/QUniBone/QUniLator - The installer can be downloaded from https://github.com/QUniBone/qunilator-installer/releases

We hope you find this software interesting to explore and are looking forward to your feedback and pull requests!

Jörg, Frits and Hans

[pbi...@gmail.com]

And studying https://qunilator.com/ ... just wow :-}!

One nit(?) is that it appears that none of the included screenshots are backed by a full-size (readable) image.  Perhaps "coming soon"?

[John H. Reinhardt]

Thanks for the announcement!  I had noticed the original GitHub repository of Joerg's going to read-only last week and did some exploring.  It looks like you have done quite a job streamlining the demo interface.  I'm looking forward to trying this out on my QBones.

One question I had was about DECromancer's QBone Dual version.  Will this run on those or has it been tested on them?  I did find it curious (and worrisome) that they weren't mentioned in the section about how to get cards.

Thanks for all your work and continued work on this!

Regards,
  John H. Reinhardt

[Chris Hanson]

On Aug 28, 2026, at 8:35 PM, Hans Hübner <hans.h...@gmail.com> wrote:
>
> We have been using Claude Code to implement everything.

Extremely disappointing to see this garbage infecting retrocomputing.

-- Chris

[Steven Hirsch]

I wasn't going to be the first to pile on, but seeing that one line erased
all interest I might have had in this project.



>
> -- Chris
>
> --
> You received this message because you are subscribed to the Google Groups "UniBone" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to unibone+u...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/unibone/A6B49410-3DCA-4801-8FAC-A901839ACD91%40gmail.com.
>

[Dave Plummer]

Cry harder. Or prove us all fools by going and writing it by hand. And to be fair, you should do it on a PDP-11/44 and a VT220.

I think it’s excellent that we’ll have these tools, presuming they’ve been adequately debugged and so on.

- Dave

[Chris Hanson]

On Aug 29, 2026, at 9:33 AM, Dave Plummer <da...@daveandnikki.com> wrote:
>
> Cry harder.

Is this what the State of Washington said when you got caught scamming computer illiterate people to make a buck?

-- Chris

[John H. Reinhardt]

On 8/29/26 11:32 AM, Steven Hirsch wrote:
> On Sat, 29 Aug 2026, Chris Hanson wrote:
>
>> On Aug 28, 2026, at 8:35 PM, Hans Hübner <hans.h...@gmail.com> wrote:
>>>
>>> We have been using Claude Code to implement everything.
>>
>> Extremely disappointing to see this garbage infecting retrocomputing.
>
> I wasn't going to be the first to pile on, but seeing that one line erased all interest I might have had in this project.
>

Here's Joerg's original code, not touched by AI. Have at it.
https://github.com/QUniBone/QUniBoneClassic

Regards,
  John H. Reinhardt

[Chris Hanson]

On Aug 28, 2026, at 8:35 PM, Hans Hübner <hans.h...@gmail.com> wrote:

 - New device emulations (TS11/TSV05, TQK50/TMSCP, KW11-P, MXV11-B2, DEUNA, VCB01/QVSS; DZV11, DHV11 and the RL/RLV12), validated against XXDP diagnostics where appliicable

 - New processor emulations: KA780 (VAX-11/780) and KD11-EA (PDP-11/34)

 - Several bug fixes

We have been using Claude Code to implement everything.

Since you mention all these new features and the use of Claude Code, how have you verified license compliance with the original codebases that it copied into your project as part of implementing these emulations?

Someone tried to submit patches to add one of the Ethernet board emulaations a while ago that contained GPL'd code copied wholesale from SIMH, presumably by using Claude Code or an equivalent LLM/genAI tool. (Since I wouldn't expect anyone to knowingly do such a thing otherwise.) How did you make sure that this didn't happen for the above?

  -- Chris

[Dave Plummer]

I wrote a book with a chapter on it. Not a secret, Chris.

Classy though!

- Dave

[Steven Hirsch]

That's what I've been using and that's what I'll continue to use.

I'm sorry folks are offended by my dislike for AI.. anything, but I feel
very strongly about machine authoring. *Review and analysis* are another
issue, but writing code? Sorry, not for me.

Every day I give thinks for being able to retire from a long career as a
software engineer just in time to miss this movement.




>
> Regards,
>   John H. Reinhardt

>
> --
> You received this message because you are subscribed to the Google Groups
> "UniBone" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to unibone+u...@googlegroups.com.
> To view this discussion visit

> https://groups.google.com/d/msgid/unibone/dfb90790-2a16-4a1f-b0ca-323d05d35c10%40thereinhardts.org.
>

[Chris Hanson]

On Aug 29, 2026, at 9:58 AM, Dave Plummer <da...@daveandnikki.com> wrote:
>
> I wrote a book with a chapter on it. Not a secret, Chris.
>
> Classy though!

Spoken like someone who has no trouble ignoring morals and ethics to get what they want, whether that's money or new features in code.

-- Chris

[Joerg Hoppe]

Guys, to be clear: this awsome outburst of new functions is solely the work of Hans and some friends ... who really know to command the AI! Kind regards, Joerg

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

To view this discussion visit https://groups.google.com/d/msgid/unibone/ECD25D2E-9FA1-45EE-A3F0-09983CC8EA27%40gmail.com.

[Mark Matlock]

John,

    I’m installing on a spare QBone Dual now. I’m doing this with a M5 MacBookPro running 26.6.2 OS and the installation is really nice!! I can say that it all installs on the QBone Dual perfectly. Now I need to test in a PDP-11.

I’ve down loaded the disk images from the main site but need to sftp over the disk images for one of my main QBus PDP-11s from the QBone running in them. The 11/73 I took to VCF SW is my experimental system and it has the other QBone Dual so I’ll try that one first. I haven’t fired it up since getting back from VCF SW because I need to replace the fans in it. I guess this gives me the motivation I need to do that.

Best,

Mark

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

To view this discussion visit https://groups.google.com/d/msgid/unibone/9a8386e6-71d0-4363-8998-9a0424fbddf8%40thereinhardts.org.

[Chris Hanson]

On Aug 29, 2026, at 10:08 AM, Joerg Hoppe <ioerg...@gmail.com> wrote:
>
> Guys, to be clear: this awsome outburst of new functions is solely the work of Hans and some friends ... who really know to command the AI! Kind regards, Joerg

If any code was inserted into the project directly by the LLM, rather than by the people involved, then that code was copied from someplace else and compliance with its copyright and license is still mandatory.

That may just mean adding notice and acknowledgement, or it may require changes to other code in the project, since not all Open Source is license-compatible with all other Open Source code. (This is why ZFS can't just be imported to the Linux kernel, for example: Sun's GPL-like license and the kernel GPLv2 license are incompatible.)

Plus there are undoubtedly proprietary codebases that LLM providers have trained on as paart of their indiscriminate scraping, since people upload things like leaked code to GitHub and other hosts all the time.

-- Chris

[Frits Jalvingh]

For all of you that feel unhappy about the route chosen to update the code: you are more than welcome to the previous version and work on it. Not meant in a nasty way, just saying that this does not take away what was there.

And I can understand your feelings about AI; it is changing my job at record speed too, and some things I took pride in can now be done by a machine, a lot faster (albeit not always better, so far).
But the Djinn is out of the bottle, and we will not get it back in. Protesting against AI and the way it is trained might increase its price - but will not change its existence, and some of the arguments like "all code it inserts is copied from somewhere" do not really hold, in truth. To each his or her opinion.

For me it is change or go under; I might not like the thing, and I am even scared for the future around these things, but for me not using them is not really an option. And they are immensely powerful as can be seen by what was done by this amazing piece of technically very complex code that was made by Joerg in a short time..

-- Frits

[Chris Hanson]

On Aug 29, 2026, at 10:44 AM, 'Frits Jalvingh' via UniBone <uni...@googlegroups.com> wrote:
>
> and some of the arguments like "all code it inserts is copied from somewhere" do not really hold, in truth.

You're very wrong about this. Or more specifically, you may be right in that the people using these tools *don't care* about where the code an LLM inserts came from (i.e. they find the argument unpersuasive). That doesn't mean the LLM isn't reproducing portions of its training set, possibly substantial portions, and thus causing the project into which that code is inserted to violate copyright and/or license.

> For me it is change or go under; I might not like the thing, and I am even scared for the future around these things, but for me not using them is not really an option. And they are immensely powerful as can be seen by what was done by this amazing piece of technically very complex code that was made by Joerg in a short time..

Your employer might require you to make moral or ethical compromises and use these tools, and as with any such compromise you will have to weigh those against your conscience and your material circumstances.

Using them *is* a moral and ethical compromise though, and if you're choosing to make such a compromise when you don't have to, such as in a hobby, you should expect to be judged for it more harshly.

-- Chris

[Frits Jalvingh]

I understand that it is your __opinion__ that I am wrong in this, Chris. I, however, do not agree that all that an AI does is copying copy-protected code. And as I do not think that this is the case I think my morality is not in question, nor my ethics.

I do agree that AI's have been trained on a lot of material that might be copyrighted. That should be resolved, and it will, hopefully keeping lawyers busy for years so that they can leave the rest of us in peace. The companies that make money with these things should, indeed, pay for what they have used - as should we, by paying for the use of AI.

As far as judging goes: you obviously can do that. But you might leave some room for people having another opinion. You state that "using them *is* a moral and ethical compromise". That is not an absolute truth but your opinion. Which is fine, but I have another one.

I will leave it at this. I do not like my ethics nor my morality questioned, so I wanted to at least address that. Please act according to your own beliefs.

Regards,
Frits

[Ethan Dicks]

On Fri, Aug 28, 2026 at 11:35 PM Hans Hübner <hans.h...@gmail.com> wrote:
> Hello,
>

> over the last couple of weeks, we have built a new version of the QBone/UniBone software...

>
> We have been using Claude Code to implement everything.

Put me in the "disappointed with this development" column.

-ethan

[Mark Matlock]

John,

    I can confirm that the new QUnilator is working perfectly with a QBone Dual. I took the SDcard that I imaged using the instructions for the Mac OS. I used SFTP to transfer the DU and DL disk images for the PDP-11/73 from the QBone Dual that was in the 11/73 to the new SDcard. Then I swapped the new card into the QBone Dual that was in the 11/73. Next I learned to use the Web based configuration page to add the DU and DL emulations, then mapped the disk images to the respective drives. To activate the emulation, one powers the configuration on in the Web dashboard, this activates the emulations and reboots the CPU. Then RSX11M+ came up perfectly.

  The user interface is very different, but MUCH easier to use. It also appears that it minimizes and likely eliminates the need for a QBone console serial connection.

Best,

Mark

On Aug 29, 2026, at 9:17 AM, John H. Reinhardt <johnhre...@thereinhardts.org> wrote:



[Mark Matlock]

Fritz, Hans, Joerg,
I just want to say that I am very impressed by these enhancements to the QUniBone project.

The new web based user interface is much easier to use!! I look forward to exploring all the new functionality and device emulations.

Thanks and Best Regards,

Mark

> On Aug 29, 2026, at 12:44 PM, 'Frits Jalvingh' via UniBone <uni...@googlegroups.com> wrote:
>
> For all of you that feel unhappy about the route chosen to update the code: you are more than welcome to the previous version and work on it. Not meant in a nasty way, just saying that this does not take away what was there.

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

> To view this discussion visit https://groups.google.com/d/msgid/unibone/rCHttG_7PDH1etzgjMkDKmnnVKo65Jw49grY_rRpHKExd_BgWdVhOGQKNqqh8mh_PbcKmvAXyfX6Y0Qv318RdIbTO_i2BQr-HI4alRbRNp0%3D%40etc.to.

[jay....@gmail.com]

On Friday, August 28, 2026 at 8:35:29 PM UTC-7 hans.h...@gmail.com wrote:

over the last couple of weeks, we have built a new version of the QBone/UniBone software. 

...

Wow! This is impressive. It even includes an MCP server!  I'll have to point my local Qwen3.8 instance at it and see what kind of trouble it can get into. :)

Thanks to Hans and others for doing this. I look forward to digging in.

On a related note, I have to say the timing of this is quite a coincidence for me, as I have also been working on a new version of the QUniBone software.  The big issue I have with the existing QUniBone images is that they are based on a full Linux distro (Debian), including systemd and all the attendant overhead that entails. While convenient to use, this makes boot times rather slow; close to a minute under optimal conditions. For me, at least, this has been a real hindrance to use, especially as it affects the hardware debug cycle. While I love the features of the new QUniLator software, I presume that the images have similar longish boot times (or maybe a little bit worse, since Debian 13 was a little slower to boot in my tests).

To address the boot time issue, I've created a new QUniBone image based on BuildRoot*.  In my version, the QUniBone software (the demo app) and the various PDP-11 disk images are largely unchanged from the existing images.  However, the host OS has been replaced with a minimal linux system image built specifically for the UniBone/QBone platforms. BuildRoot uses minimal system utilities, and a traditional rc boot script that greatly streamlines the boot process. I've also disabled a number of unnecessary kernel features which helps with time spent during initialization. Currently, boot times are under 8 seconds (not counting network address acquisition), which is a much more comfortable wait time for me. All existing QUniBone features work as they do on the original images, and the new images retain most of the supporting software (bash, ssh, perl, python, emacs, vim, etc.) plus some new PDP-11 stuff (simh pdp11, macro11, pdp11monloader, xxdpdir).

Now, with this new QUniLator software it looks like I'll have to regroup a bit and see what it takes to port this on top of BuildRoot. I'm hopeful it will be a straightforward process.

Thanks again for all the great work.

--Jay

* -- Apropos of the meta-discussion, this was not written by an AI, although I'm not against using them in certain situations.

[Steven Hirsch]

On Sat, 29 Aug 2026, jay....@gmail.com wrote:

> To address the boot time issue, I've created a new QUniBone image based
> on BuildRoot*.  In my version, the QUniBone software (the demo app) and
> the various PDP-11 disk images are largely unchanged from the existing
> images.  However, the host OS has been replaced with a minimal linux
> system image built specifically for the UniBone/QBone platforms.
> BuildRoot uses minimal system utilities, and a traditional rc boot
> script that greatly streamlines the boot process. I've also disabled a
> number of unnecessary kernel features which helps with time spent during
> initialization. Currently, boot times are under 8 seconds (not counting
> network address acquisition), which is a much more comfortable wait time
> for me. All existing QUniBone features work as they do on the original
> images, and the new images retain most of the supporting software (bash,
> ssh, perl, python, emacs, vim, etc.) plus some new PDP-11 stuff (simh
> pdp11, macro11, pdp11monloader, xxdpdir).

I'd be interesting trying out this image when you feel it's stable enough
for public consumption! Sounds like great work.

[Hans Hübner]

Am Sa., 29. Aug. 2026 um 08:30 Uhr schrieb pbi...@gmail.com <pbi...@gmail.com>:

One nit(?) is that it appears that none of the included screenshots are backed by a full-size (readable) image.  Perhaps "coming soon"?

I've updated the manual so that clicking images opens a higher-res version. Thank you for the feedback! 

[pbi...@gmail.com]

The higher-res versions are quite effective :-}.  Thank you for the update.