From Newsgroup: alt.comp.os.windows-11

On Thu, 6/11/2026 4:03 AM, T wrote:

Hi All,

When you download the iso form:
-a-a-a https://www.microsoft.com/en-us/software-download/windows11

You get:
-a-a-a Win11_25H2_English_x64_v2.iso

But when you visit:
-a https://learn.microsoft.com/en-us/windows/release-health/windows11-release-information

It states:
-a-a-a Windows 11, version 26H1 is scoped to support new devices that
-a-a-a come to market in early 2026 and is not designed as a feature
-a-a-a update for existing devices. This version is not offered as
-a-a-a an in-place update from 24H2 or 25H2 on existing devices.

-a-a-a Version-a-a-a-a Servicing option-a-a-a-a-a-a-a-a-a-a-a Availability date
-a-a-a 26H1-a-a-a-a-a-a-a General Availability Channel-a-a-a 2026-02-10

Hmmmmmmmm ....
What about new 2026 hardware ???
Is 26H1 top secret?

Yours in Confusion,
-T

The 26H1 is for Qualcomm ARM machines.
It's a separate stream for hardware.

It's possible something like a DGX sparc could use
the same stream (it could be an ARM core as well).

More of the machines would have x86_64 processors
and use 25H2. Which is what a lot of people would be
using here.

You wouldn't want x86 and ARM native, on the same DVD
as the DVD would be the size of a hard drive :-) The
DVD is big enough as it is.

Paul


If you're on an ARM machine, there is some sort of setup
for interpreting x86 code. Whether it is both 32 bit and
64 bit code, I don't know.
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 6/11/26 6:09 AM, Paul wrote:

On Thu, 6/11/2026 4:03 AM, T wrote:

Hi All,

When you download the iso form:
-a-a-a https://www.microsoft.com/en-us/software-download/windows11

You get:
-a-a-a Win11_25H2_English_x64_v2.iso

But when you visit:

https://learn.microsoft.com/en-us/windows/release-health/windows11-release-information

It states:
-a-a-a Windows 11, version 26H1 is scoped to support new devices that
-a-a-a come to market in early 2026 and is not designed as a feature
-a-a-a update for existing devices. This version is not offered as
-a-a-a an in-place update from 24H2 or 25H2 on existing devices.

-a-a-a Version-a-a-a-a Servicing option-a-a-a-a-a-a-a-a-a-a-a Availability date
-a-a-a 26H1-a-a-a-a-a-a-a General Availability Channel-a-a-a 2026-02-10


Hmmmmmmmm ....
What about new 2026 hardware ???
Is 26H1 top secret?


Yours in Confusion,
-T

The 26H1 is for Qualcomm ARM machines.
It's a separate stream for hardware.

It's possible something like a DGX sparc could use
the same stream (it could be an ARM core as well).

More of the machines would have x86_64 processors
and use 25H2. Which is what a lot of people would be
using here.

You wouldn't want x86 and ARM native, on the same DVD
as the DVD would be the size of a hard drive :-) The
DVD is big enough as it is.

Paul

If you're on an ARM machine, there is some sort of setup
for interpreting x86 code. Whether it is both 32 bit and
64 bit code, I don't know.

Hi Paul,

Not to ask too stupid a question , but will x86_64
programs run on 26H1?

Yours in Confusion,
-T
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On Sat, 6/13/2026 3:52 AM, T wrote:

On 6/11/26 6:09 AM, Paul wrote:

On Thu, 6/11/2026 4:03 AM, T wrote:

Hi All,

When you download the iso form:
-a-a-a-a https://www.microsoft.com/en-us/software-download/windows11

You get:
-a-a-a-a Win11_25H2_English_x64_v2.iso

But when you visit:
-a https://learn.microsoft.com/en-us/windows/release-health/windows11-release-information

It states:
-a-a-a-a Windows 11, version 26H1 is scoped to support new devices that
-a-a-a-a come to market in early 2026 and is not designed as a feature
-a-a-a-a update for existing devices. This version is not offered as
-a-a-a-a an in-place update from 24H2 or 25H2 on existing devices.

-a-a-a-a Version-a-a-a-a Servicing option-a-a-a-a-a-a-a-a-a-a-a Availability date
-a-a-a-a 26H1-a-a-a-a-a-a-a General Availability Channel-a-a-a 2026-02-10 >>>

Hmmmmmmmm ....
What about new 2026 hardware ???
Is 26H1 top secret?


Yours in Confusion,
-T

The 26H1 is for Qualcomm ARM machines.
It's a separate stream for hardware.

It's possible something like a DGX sparc could use
the same stream (it could be an ARM core as well).

More of the machines would have x86_64 processors
and use 25H2. Which is what a lot of people would be
using here.

You wouldn't want x86 and ARM native, on the same DVD
as the DVD would be the size of a hard drive :-) The
DVD is big enough as it is.

-a-a-a Paul


If you're on an ARM machine, there is some sort of setup
for interpreting x86 code. Whether it is both 32 bit and
64 bit code, I don't know.

Hi Paul,

Not to ask too stupid a question , but will x86_64
programs run on 26H1?

Yours in Confusion,
-T

Previously, Intel threatened to sue, if Microsoft did a x64 to ARM translator. According to this, the W11 one does both.

https://www.reddit.com/r/hardware/comments/1bg4620/how_does_x86_translation_work_on_windows_on_arm/

"It's actually a JIT translator and saves the translated code cache."

"Win10 arm emulation supported only 32 bit apps "
"Yes, Win11 emulation supports both 32 and 64 bit"

As Microsoft tried ARM before, it is not clear what the
justification for it is this time. The people buying the Qualcomm
product, were doing it because of battery life, not because it
had an NPU or something. And if you look at recent Intel offerings,
at least some of them have 13 TOPs NPU and not the 50 TOPs a few had.
Intel is not riding the NPU-wagon right now. You would think
making all NPU offerings 50 TOPS would make sense.

It's not like the industry has a plan right now, anyway. Intel
will be on the look-out for "things to fill its fabs". You have to
run a fab at 95% load, to stay in business. IBM got rid of its
fabs, because the number of wafers they were running had dropped,
and their CMOS process (we used to use it), just wasn't creating
the fab traffic. Well, Intel runs from the same enemy, the empty fab.
the fab is very expensive, and it must constantly make money... to
pay for the next 10 billion dollar fab. This is why Intel has been
trying to sell fab services to others, and may have snagged some chip-plans which will be pipe cleaners for their fab offerings to others. At least
one deal they've got is "chip packaging", and that does not particularly
work to their favor (it doesn't keep a fab line full, it competes with
their packaging needs for their server CPUs).

We used to think, that "Intel had the edge, due to defining its own CMOS processes".
But that has proved to not be true. Others have been able to stay within inch-worm distance of Intel or vice versa, so there is no clock rate
advantage particularly. The 6GHz 14900K turbo was the last bit of gee-whiz
they had, and their microcode-level voltage booster was ruining the 13 and 14 gen processors, so they had to crank that down to save their reputation.
That's the only reason that buying a 14900K today, is safe, is BIOS updates with the microcode thing turned down.

But whether anyone can "launch" anything, that's unclear to me. Like, so
what if DGX Sparc was released, if there wasn't enough RAM to solder to
one of those, to make it to market. There could be a lot of "soft launches" where there isn't RAM to launch equipment properly. And the plan last week,
was to make laptops with 8GB of RAM (basically, a single DIMM or SODIMM).
If you were thinking in terms of machines with 128GB, 192GB, 256GB as some
of the product announcements indicated, maybe they can make three of them before they go "out of stock" :-)

Some of Intel and AMD announcements, are plays on older RAM standards.
There is even a plan to re-introduce 3060 video cards back to the vid card market,
just because the GDDR standard those use, there might be some chips sitting
in a drawer somewhere.

With all that is going on, it's hard to see how any "technically superior" ideas
can gain any traction, when every product is in danger of going "out of stock". At the computer store, the look on the customers face when the clerk
told him how much the store DDR5 cost (none on display...) and the scowl
on the customers face. The customer was really about to pull the trigger
on a home build, and the RAM details stopped him in his tracks. The clerk finished his pricing summary with "Do you need financing?", which of course really irritates the people who think they are "rich cats".

The question then, is what kind of interest would there be in an
ARM offering with a single 8GB stick and no NPU. To the customer,
the only thing it would have going for it is battery life. And it
had better have a damn fine keyboard for the note-taker people who
need that battery life. If it had a big NPU or a GPU with a lot of
execution units, then the customer expects 128GB minimum RAM. Which for
most of those customers, isn't going to happen.

Paul
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 06/13/2026 4:47 AM, Paul wrote:

On Sat, 6/13/2026 3:52 AM, T wrote:

On 6/11/26 6:09 AM, Paul wrote:

On Thu, 6/11/2026 4:03 AM, T wrote:

Hi All,

When you download the iso form:
-a-a-a-a https://www.microsoft.com/en-us/software-download/windows11

You get:
-a-a-a-a Win11_25H2_English_x64_v2.iso

But when you visit:
-a https://learn.microsoft.com/en-us/windows/release-health/windows11-release-information

It states:
-a-a-a-a Windows 11, version 26H1 is scoped to support new devices that >>>> -a-a-a-a come to market in early 2026 and is not designed as a feature >>>> -a-a-a-a update for existing devices. This version is not offered as
-a-a-a-a an in-place update from 24H2 or 25H2 on existing devices.

-a-a-a-a Version-a-a-a-a Servicing option-a-a-a-a-a-a-a-a-a-a-a Availability date
-a-a-a-a 26H1-a-a-a-a-a-a-a General Availability Channel-a-a-a 2026-02-10 >>>>

Hmmmmmmmm ....
What about new 2026 hardware ???
Is 26H1 top secret?


Yours in Confusion,
-T

The 26H1 is for Qualcomm ARM machines.
It's a separate stream for hardware.

It's possible something like a DGX sparc could use
the same stream (it could be an ARM core as well).

More of the machines would have x86_64 processors
and use 25H2. Which is what a lot of people would be
using here.

You wouldn't want x86 and ARM native, on the same DVD
as the DVD would be the size of a hard drive :-) The
DVD is big enough as it is.

-a-a-a Paul


If you're on an ARM machine, there is some sort of setup
for interpreting x86 code. Whether it is both 32 bit and
64 bit code, I don't know.

Hi Paul,

Not to ask too stupid a question , but will x86_64
programs run on 26H1?

Yours in Confusion,
-T

Previously, Intel threatened to sue, if Microsoft did a x64 to ARM translator.
According to this, the W11 one does both.

https://www.reddit.com/r/hardware/comments/1bg4620/how_does_x86_translation_work_on_windows_on_arm/

"It's actually a JIT translator and saves the translated code cache."

"Win10 arm emulation supported only 32 bit apps "
"Yes, Win11 emulation supports both 32 and 64 bit"

As Microsoft tried ARM before, it is not clear what the
justification for it is this time. The people buying the Qualcomm
product, were doing it because of battery life, not because it
had an NPU or something. And if you look at recent Intel offerings,
at least some of them have 13 TOPs NPU and not the 50 TOPs a few had.
Intel is not riding the NPU-wagon right now. You would think
making all NPU offerings 50 TOPS would make sense.

Paul

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.
--
...w-i|#-o-#-n|#
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On Sat, 6/13/2026 11:52 AM, ....winston wrote:

The short answer is no.
-a- not deployed to x86-x64 machines.
-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps)
-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a- 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 ?

*******

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. Which
means I'd have to check all the URLS -- and some recently included URLs
in answers, were AI slop web pages, and that really annoys me as it
encourages ingestion and circularization of (questionable) information.
The usage of a lot of BOLD in the mundobytes article, suggests the
origin might be LLM-AI based.

*********************** CoPilot Question ********************************

This is a question about Microsoft PRISM code translation on ARM-based computers (Windows 11).

What types of executables that can be translated ?

Is there a provision for Metro.Apps and legacy 32-bit PE and can 64-bit PE+ be translated ?

*********************** CoPilot Answer ********************************

Short answer:

On WindowsrC>11 ARM with MicrosoftrCOs Prism translation layer,

Win32 x86 (32-bit) and Win32 x64 (64-bit) PE/PE+ executables are translatable,

while UWP/Metro apps and anything requiring kernelrCamode drivers are not.

Below is the precise breakdown, grounded in current Microsoft documentation and technical reports. <=== Cites ??? Wally ???

---

## **1. What types of executables Prism can translate**
Prism translates **userrCamode x86 and x64 code** into Arm64 at runtime. This includes:

- **32rCabit x86 PE executables** (classic Win32 apps)
https://mundobytes.com/en/Installing-x86-and-x64-apps-on-Windows-11-ARM-with-Prism%3A-Complete-compatibility-guide--limits--and-workarounds/

- **64rCabit x64 PE+ executables**
Windows 11 24H2 added expanded x64 support, including AVX/AVX2, BMI, FMA, F16C instruction emulation.
https://www.elevenforum.com/t/change-windows-on-arm-emulation-settings-for-app-in-windows-11.41551/

- **x86/x64 DLLs** loaded into those processes
(Prism translates all userrCamode modules in the process.)
https://mundobytes.com/en/Installing-x86-and-x64-apps-on-Windows-11-ARM-with-Prism%3A-Complete-compatibility-guide--limits--and-workarounds/

Prism uses JIT translation with caching, so subsequent launches reuse previously translated blocks.
https://mundobytes.com/en/Installing-x86-and-x64-apps-on-Windows-11-ARM-with-Prism%3A-Complete-compatibility-guide--limits--and-workarounds/

## **2. What Prism *cannot* translate**
### **A. KernelrCamode components**
Prism does **not** emulate kernelrCamode code:

- Drivers (KMDF, WDM)
- File system filters
- Antivirus kernel hooks
- Virtualization engines requiring kernel components

All kernelrCamode drivers **must be Arm64-native**.
https://www.pchardwarepro.com/en/How-to-troubleshoot-x86-application-compatibility-issues-on-Windows-11-for-ARM/

If an app depends on a non-Arm64 driver, it will not run.

---

### **B. UWP / Metro / Microsoft Store apps**
This is the part most people miss:

- **Microsoft Store apps (UWP/Metro)** do **not** use Prism emulation settings
(Microsoft explicitly states emulation settings do not apply to Store apps).
https://www.elevenforum.com/t/change-windows-on-arm-emulation-settings-for-app-in-windows-11.41551/

Why?
Because UWP apps are packaged with architecture-specific binaries. If the Store app does
not include an Arm64 or x86/x64 package compatible with the device, Windows simply refuses
to install it. There is **no translation path** for UWP binaries.

So:

| App Type | Translated? | Notes |
|-----------------------------------------------|-------------|----------------------------------------|
| **Win32 x86** | Yes | Full Prism support |
| **Win32 x64** | Yes | Expanded CPU feature emulation in 24H2 |
| **UWP/Metro apps** | No | Must include native Arm64 package |
| **Drivers** | No | Must be Arm64-native |
| **Shell extensions, codecs, browser plugins** | No | Must be Arm64-native |

---

## **3. Can 64rCabit PE+ executables be translated?**
**Yes.**
Windows 11 ARM (24H2 and later) supports **x64 emulation** through Prism, including:

- x64 userrCamode binaries
- x64 instruction sets including AVX/AVX2, BMI, FMA, F16C (new in 24H2)
https://www.elevenforum.com/t/change-windows-on-arm-emulation-settings-for-app-in-windows-11.41551/

This is a major improvement over Windows 10 ARM, which only supported x86 (32rCabit) emulation.

---

## **4. What about legacy 32rCabit PE?**
**Fully supported.**
WOW64 is present on Windows ARM and handles x86 processes, with Prism performing the translation.
https://mundobytes.com/en/Installing-x86-and-x64-apps-on-Windows-11-ARM-with-Prism%3A-Complete-compatibility-guide--limits--and-workarounds/

---

## **5. What about Metro/UWP apps?**
To be explicit:

- **There is no translation path for UWP/Metro apps.**
- They must ship with an **Arm64 package**.
- If they only ship x86/x64, the Microsoft Store will not install them.
- Emulation settings do not apply to them.
https://www.elevenforum.com/t/change-windows-on-arm-emulation-settings-for-app-in-windows-11.41551/

---

## **Summary Table**

| Executable Type | Prism Translation? | Notes |
|---------------------------------------------|--------------------|------------------------------------------------------------|
| **x86 (32-bit) PE** | Yes | Full support; WOW64 + Prism |
| **x64 (64-bit) PE+** | Yes | Full support; expanded CPU features in 24H2 |
| **UWP/Metro apps** | No | Must include Arm64 package; Store blocks incompatible ones |
| **Kernel-mode drivers** | No | Must be Arm64-native |
| **Shell extensions / codecs / COM objects** | No | Must be Arm64-native |
| **Arm32 (deprecated)** | Limited | Microsoft is phasing out ARM32 support |

---

## If you want, I can also explain:
- How Arm64EC fits into this picture (hybrid native/emulated processes)
- How Prism caches translated blocks and how CHPE changed in 24H2
- How to inspect a PE file to see if it will run under Prism
- Why some installers fail even though the app itself would run

Just tell me what direction you want to go.
*********************** CoPilot Answer ********************************

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.

Paul
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 06/13/2026 12:59 PM, Paul wrote:

On Sat, 6/13/2026 11:52 AM, ....winston wrote:

The short answer is no.
-a- not deployed to x86-x64 machines.
-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps)
-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a- 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.

Paul

--
...w-i|#-o-#-n|#
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

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.
-a-a- not deployed to x86-x64 machines.
-a-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps)
-a-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a-a- 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.

-aWell, 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.

A better approach is to set the AI to "high reasoning" (also
known as simulated reasoning), as then the LLM-AI can examine the
convergence properties of each of a series of question-runs, and from
that it can determine whether the answers have poor math properties.

We tried that on Micky's question a few weeks ago, and the LLM-AI eventually admitted (without prompting, without grinding its beak into the ground),
that the "simulated reasoning" it had done told it that what it was doing
was "guessing". The LLM-AI refuse to admit that, on single-run sessions,
such as the one I am quoting above.

But what I do find noteworthy, is there are no citations of Microsoft
web pages with PRISM information. And as far as I know, PRISM has been
around for a while.

*******

We are forced to learn how to use these things, by experiment. And
develop our own mental models on whether anything we do smacks of
"control".

You cannot tell it to do certain things in plain language. For
example "use high reasoning" or "use simulated reasoning" is
out of the question. You can try "work slowly and carefully,
giving a precise answer". In other words, you don't order it around,
but you describe (for the benefit of the strategy planner), the
"flavor" of answer you want, and then you might get the quality
of answer you were after.

On the first problem we studied here, the list the US presidents
and put the answer in a certain order, it took a
"And don't lose any of the Presidents!" statement to cause
it to stop discarding one particular president. This would
be a non-controversial individual, where it would be hard to
explain why that individual kept getting dropped. We were also
able to find web citations of people doing variants of that
experiment and noticing the same thing. The same president
kept going missing. Why would we invest trillions of dollars,
in something this idiotic ??? It's like a Magic 8 Ball.

Paul
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 06/13/2026 10:18 PM, 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.
-a-a- not deployed to x86-x64 machines.
-a-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps) >>>> -a-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a-a- 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.

-aWell, 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.

A better approach is to set the AI to "high reasoning" (also
known as simulated reasoning), as then the LLM-AI can examine the
convergence properties of each of a series of question-runs, and from
that it can determine whether the answers have poor math properties.

We tried that on Micky's question a few weeks ago, and the LLM-AI eventually admitted (without prompting, without grinding its beak into the ground),
that the "simulated reasoning" it had done told it that what it was doing
was "guessing". The LLM-AI refuse to admit that, on single-run sessions,
such as the one I am quoting above.

:) It(AI), in your Micky reference is akin to questioning or reason why
the Egyptian pharaoh's built pyramids for tombs.
No source exists - anything/anyone trying to answer same is pure
supposition and postulation and lacking any method to validate.

AI would probably tell you something like reflective of beliefs about
royal standing, afterlife, religion and politics(power).
- Pure supposition again and likely from no written, documented source
from that point in time.

But what I do find noteworthy, is there are no citations of Microsoft
web pages with PRISM information. And as far as I know, PRISM has been
around for a while.

Prism was released as part of Windows 11 24H2 on ARM devices in June
2024, not eliminating(or deprecating) but improving upon existing
inferior emulation performance present in older Windows on ARM technology.

At this stage, Prism's purpose, as you know(or learned more from AI) is specific to run x86/64 applications natively on Windows on ARM devices
by translating x86/64 code/instructions into ARM code(on the fly while
the application is running). Also provides the ability to install(and
open) x86/64 supported programs on an ARM device without user direction
or control to manage emulation.

*******

We are forced to learn how to use these things, by experiment. And
develop our own mental models on whether anything we do smacks of
"control".

You cannot tell it to do certain things in plain language. For
example "use high reasoning" or "use simulated reasoning" is
out of the question. You can try "work slowly and carefully,
giving a precise answer". In other words, you don't order it around,
but you describe (for the benefit of the strategy planner), the
"flavor" of answer you want, and then you might get the quality
of answer you were after.

Welcome to the machine, and the reality that infallibility does not
exist in humans, logic or science, though some politicians might lead
you believe their rhetoric in any form is!

On the first problem we studied here, the list the US presidents
and put the answer in a certain order, it took a
"And don't lose any of the Presidents!" statement to cause
it to stop discarding one particular president. This would
be a non-controversial individual, where it would be hard to
explain why that individual kept getting dropped. We were also
able to find web citations of people doing variants of that
experiment and noticing the same thing. The same president
kept going missing. Why would we invest trillions of dollars,
in something this idiotic ??? It's like a Magic 8 Ball.

See above <g>
--
...w-i|#-o-#-n|#
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

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.
-a-a- not deployed to x86-x64 machines.
-a-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps)
-a-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a-a- 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.

-aWell, 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/

One of the reasons I left the Apple Ecosystem, was the running of mixed-this-and-that
all the time. So if someone were to ask me ""whether a W11 ARM laptop was the equal of a W11 x86 laptop", the answer would be "well, how much of a learning experience
do you want". I don't particularly care that *any* computer is 5% faster, I would
not trade actual seamless operation for "odd surprises early in the morning" where
a project is scuttled because a wall of tick boxes was in the path. For example, if
I bought an Apple M5, the very next day they would announce that everything
was being changed to RISCV.

All the games I played on Apple computers, ran at 20FPS. It was a manifest constant
of the ecosystem, like the gravitational constant of the Moon. Today, I don't have to
worry about that any more.

Paul


--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

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.
-a-a- not deployed to x86-x64 machines.
-a-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps) >>>> -a-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a-a- 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.

-aWell, 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.
--
...w-i|#-o-#-n|#
--- Synchronet 3.22a-Linux NewsLink 1.2

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.
-a-a-a- not deployed to x86-x64 machines.
-a-a-a- Prism emulation on ARM can run most x86-64 apps(keyword = apps) >>>>> -a-a-a- 26H1 targets new silicon(i.e. adoption of devices with silicon that requires[it])
-a-a-a- 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.
-a-a-> 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.

-a-aWell, 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". >>
-a-a-a-a 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.

-a-a-a-a 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/

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

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


-a-a-a 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 3.22a-Linux NewsLink 1.2

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-i|#-o-#-n|#
--- Synchronet 3.22a-Linux NewsLink 1.2