• Re: Win11_25H2_English_x64_v2.iso ???

    From Paul@nospam@needed.invalid to alt.comp.os.windows-11 on Mon Jun 15 06:32:55 2026
    From Newsgroup: alt.comp.os.windows-11

    On Sun, 6/14/2026 11:21 AM, ....winston wrote:
    On 06/14/2026 5:06 AM, Paul wrote:
    On Sat, 6/13/2026 9:08 PM, ....winston wrote:
    On 06/13/2026 12:59 PM, Paul wrote:
    On Sat, 6/13/2026 11:52 AM, ....winston wrote:


    The short answer is no.
       - not deployed to x86-x64 machines.
       - Prism emulation on ARM can run most x86-64 apps(keyword = apps) >>>>>    - 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
       - 26H1 internally(and select Enterprise partners) is available for testing or evaluation of Enterprise editions.

    The NVidia machine then, will it be 26H1 DVD and a bunch of updates ?
    Or will there be an updated version of 26H1 DVD with enough drivers
    for the DGX ?

    No DVD's. No public download.
      -> pre-install only

    *******

    I figured this would be a good CoPilot question. I don't really know how >>>> much of the topic would be in the training set.

    What is curious about this answer, is no Microsoft URLs are included.
    This means we're relying on some parr-boiled answers off websites.
    If we learn anything from the info in that answer, it's that there are a lot
    of cases we didn't think about, or for that matter, know about. The mundobytes
    article claims that some installers do not get translated, in order for a program
    install to finish.
      Well, AI has only shown it's inability to provide actual and validated source(MSFT) information, i.e. relying on external information with interpretive(potential, but not validated) responses.

    A better AI approach might be to start with a 'learn' mode[Concept] not initially asking for a response, but a staged approach of it asking the questions.

    Don't set your hope's too high...and if AI goes into 'term-paper' response mode, it's time to place everything in the circular file.

    One of the problems with the topic, is the language used in
    each article differs.

    The first one addresses Prism and uses terms like "x86 apps" and "x64 apps". >>
         https://learn.microsoft.com/en-us/windows/arm/apps-on-arm-x86-emulation

    The article on WOW64 talks of 32-bit applications, which I take to mean a PE EXE.

         https://learn.microsoft.com/en-us/windows/win32/winprog64/running-32-bit-applications

    https://www.elevenforum.com/t/change-windows-on-arm-emulation-settings-for-app-in-windows-11.41551/

         # Well, now I'm having acid flashbacks, to a time when I was dealing with
         # a mixed environment of executables on a computer. Thank goodness for tick
         # boxes and a new hobby.

         https://www.elevenforum.com/attachments/windows_on_arm_emulator-2-webp.151158/


        Paul



    I wouldn't over analyze it.

    Both Wow64 and Prism provide emulation allowing older applications to run on newer hardware.
    Prism => emulates x86/x64 on 'ARM' processors
    WOW64 => emulates 32 bit x86 on '64-bit' x86 processors.

    We have to understand it, so we can answer peoples questions.

    I have to analyze every LLM-AI question I ask, in terms of how
    can I improve the result the next time. As it comes with no
    instructions. And whereas early on, I was hearing things
    like "simulation temperature" and "simulated annealing" as
    terminologies (these might be settings in the research lab),
    you no longer see references like that at all. Not every response
    involves "hallucination", there are specific reasons for that to happen.
    But then, they won't tell us the settings they are using, for the
    various Mixture-Of-Experts. You can make a <cough> math expert out
    of a regular LLM-AI, by "turning up some settings". Yet, all this
    seems to do, is corrupt what it is doing, and give weird results.
    I'm only explaining this to you, as some more expensive models
    are quite capable of doing math (as in theoretical math), but we
    don't get to use those for free. Those can at least count.

    One product I did a brief research on, was claimed to be an "image oriented" LLM-AI.
    It was claimed to do OCR. Well, it does and it doesn't. I reads cash register slips (exactly why this is important, I have yet to find out). It reads maybe a row or two of the cash register slip at a time. What's fancy about it, is there can be emoji or icons or the like, which it can attempt to render as a text. That's good. But as far as converting a legal document scan into a PDF,
    it can't really do that. Even cheesy commercial OCR do a better job. For that product
    then, I have to wait for a more impressive demo.

    When you look at the OCR which the W11 Snippingtool does, it's both
    impressive and not so impressive. I used my digital camera in video mode,
    to record the boot screen during which the Secure Boot prints the PCR7
    status as a sentence on the screen. You cannot press the Break key and
    stop the screen from updating, so only shooting video allows recording
    what it says. My camera was on an angle, the text crooked and distorted.
    I fed a sentence or two of the screen to Snippingtool, and it got that
    sentence 100% correct. Even though the signal to noise was poor and
    any ordinary OCR tool would have tipped over and made garbage.

    Yet, the same tool is not capable of recognizing popular layouts for
    documents, in the way that the Adobe Distiller OCR in the year 2001 could do. For all of the fancy tech behind these things, they still have not surpassed some of the procedural programming from 20 years ago. That's one thing
    to remember, the next time someone shows you a half baked demo.

    Paul
    --- Synchronet 3.21d-Linux NewsLink 1.2
  • From ....winston@winstonmvp@gmail.com to alt.comp.os.windows-11 on Wed Jun 17 12:34:15 2026
    From Newsgroup: alt.comp.os.windows-11

    On 06/15/2026 6:32 AM, Paul wrote:
    On Sun, 6/14/2026 11:21 AM, ....winston wrote:
    I wouldn't over analyze it.

    Both Wow64 and Prism provide emulation allowing older applications to run on newer hardware.
    Prism => emulates x86/x64 on 'ARM' processors
    WOW64 => emulates 32 bit x86 on '64-bit' x86 processors.

    We have to understand it, so we can answer peoples questions.

    Agreed, but with limitations. Tribal knowledge and verified, validated source(provider, developer) content is not always available..thus iirc
    noted earlier our(anyone's) understanding can be skewed by supposition/opinion.

    I have to analyze every LLM-AI question I ask, in terms of how
    can I improve the result the next time. As it comes with no
    instructions. And whereas early on, I was hearing things
    like "simulation temperature" and "simulated annealing" as
    terminologies (these might be settings in the research lab),
    you no longer see references like that at all. Not every response
    involves "hallucination", there are specific reasons for that to happen.
    But then, they won't tell us the settings they are using, for the
    various Mixture-Of-Experts. You can make a <cough> math expert out
    of a regular LLM-AI, by "turning up some settings". Yet, all this
    seems to do, is corrupt what it is doing, and give weird results.
    I'm only explaining this to you, as some more expensive models
    are quite capable of doing math (as in theoretical math), but we
    don't get to use those for free. Those can at least count.


    I struggle with AI and LLM-AI(ChatGpt, Gemini, etc.) for many topics
    since results are based on 'what can be found from anywhere' to predict
    some form of digested analysis arranged with the appearance of ordered summation indicating accuracy...which as you indirectly inferred can
    output incorrect content(i.e. hallucination) and more bias(based on
    those skewed supposition/opinions).

    Using AI in any form(including LLM) can be digressive...just like this thread(Win11 25H2 Iso, ARM, X86-x64, Prism, 26H1, AI-LLM) which may
    prove useful for answering future 'peoples' questions or we could just
    be adding more hallucination for the rest of the world to consider and
    further digress.
    --
    ...w¡ñ§±¤ñ
    --- Synchronet 3.21d-Linux NewsLink 1.2