Without Linux, there is no ChatGPT. No AI at all. None. Here's why.

Modern AI began with open source, and it ran on Linux. Today, Linux isn't just important for artificial intelligence; it's the foundation upon which today's entire modern AI stack runs. From hyperscale training clusters down to edge inference boxes, it's all Linux from top to bottom.

AI's magic tricks are really the aggregate output of very prosaic infrastructure: supercomputers, GPU farms, and cloud clusters that almost all run some flavor of Linux. The core machine-learning frameworks -- TensorFlow, PyTorch, scikit-learn, and friends -- were all developed and tuned first on Linux. Tooling around these tools, from Jupyter and Anaconda to Docker and Kubernetes, is similarly optimized for Linux.

Why IT jobs will live and die on Linux

Why? Because it's on Linux where researchers and production engineers actually deploy AI. Future IT jobs will live and die on Linux.

You see, AI runs on Linux because it's the most flexible, powerful, and scalable environment for the GPU‑heavy, distributed workloads modern AI requires. In addition, the entire tooling and cloud ecosystem has standardized on Linux.

Yes, every AI platform, whether it's OpenAI, Copilot, Perplexity, Anthropic, or your favorite AI chatbot, is built on Linux, plus drivers, libraries, and orchestration, all glued together in different ways. The proprietary bits may grab the branding, but without Linux, they're nowhere.

That translates into more Linux jobs.

As the Linux Foundation's 2025 State of Tech Talent Report noted, AI is driving a net increase in tech jobs, particularly Linux jobs. What this looks like comes down to "AI [is] reshaping roles rather than eliminating them," according to the report, "leading to shifts in skill demand and new opportunities for workforce growth."

Besides increasing Linux system and network administration jobs, the site Linux Careers sees "a rapidly emerging trend involving professionals who combine Linux expertise with artificial intelligence and machine learning operations." Such new AI/Linux jobs include AI Operations Specialist, MLOps Engineer, ML Engineer, and DevOps/AI Engineer.

Of course, Linux distributors know all this, which is why, when new Linux distros are released, their makers emphasize AI features.

For example, Canonical and Red Hat are racing to plant their Linux flags on Nvidia's new Vera Rubin AI supercomputer platform. The race is on to see who will own the operating system layer of "gigascale AI factories."

For its part, Red Hat is introducing Red Hat Enterprise Linux (RHEL) for Nvidia. This curated edition of RHEL is optimized specifically for Nvidia's Rubin platform, including the Vera Rubin NVL72 rack-scale systems.

The company says this variant will ship with Day 0 support for the Vera CPU, Rubin GPUs, and Nvidia's CUDA X stack, with validated OpenRM drivers and toolkits delivered directly through Red Hat repositories.

The Linux kernel and AI

Canonical is also rolling out official Ubuntu support for the Nvidia Rubin platform, also targeting the Vera Rubin NVL72. The London-headquartered company is anchoring its story around making the custom Arm-based Vera CPU a "first-class citizen," with x86 parity in its forthcoming Ubuntu 26.04 release.

So, unlike Red Hat, which has a RHEL just for Nvidia's processors, the new Ubuntu will support Nvidia. This version will also upstream features such as Nested Virtualization and ARM Memory Partitioning and Monitoring (MPAM) to better partition memory bandwidth and cache for multi-tenant AI workloads.

What runs all this is a Linux kernel that has been steadily modified to keep up with AI's voracious appetite for hardware acceleration. Modern kernels juggle GPU and specialized accelerator drivers, sophisticated memory management for moving tensors around quickly, and schedulers tuned for massively parallel batch jobs.

In short, the kernel has been rewired over the last decade to become an operating system for AI hardware accelerators.

Memory: putting data where the GPUs are

Specifically, one of the most important enablers has been Heterogeneous Memory Management. This enables device memory, such as Graphics Processing Unit/Video Random Access Memory (GPU VRAM), to be integrated into Linux's virtual memory subsystem.

That, combined with Direct Memory Access Buffering (DMA-BUF) and Non-Uniform Memory Access (NUMA) optimization, enables AI runtimes to keep tensors close to the accelerator and cut back on data copying, which tends to slow down performance.

Recent kernels also treat advanced CPU-GPU combinations, such as tightly coupled NUMA-style CPU/GPU nodes, as first-class citizens. With this, memory can be migrated between CPU-attached RAM and high-bandwidth GPU memory on demand.

This, as Nvidia explained, "enables the CPU and GPU to share a single per-process page table, enabling all CPU and GPU threads to access all system-allocated memory."

Accelerators: a real subsystem, not an add-on

Linux now has a dedicated compute accelerators subsystem that's designed to expose GPUs, Tensor Processing Units (TPUs), and custom AI application-specific integrated circuits (ASICs) to your AI and machine learning (ML) programs.

On top of that, GPU support has matured from graphics-first to compute-heavy, via the Direct Rendering Manager (DRM), open stacks like ROCm and OpenCL, and Nvidia's Compute Unified Device Architecture (CUDA) drivers.

Kernel work has expanded to cover newer AI accelerators such as Intel's Habana Gaudi, Google's Edge TPU, and FPGA/ASIC boards, with drivers and bus abstractions. This enables AI programs such as PyTorch or TensorFlow to see and use them as just another device. Thus, anyone making new AI silicon today rightly assumes that Linux will be running on it.

Scheduling: feeding hungry accelerators

Linux's default scheduler, the Earliest Eligible Virtual Deadline First (EEVDF), real-time scheduler, and NUMA balancing have all been tuned to enable AI workloads to pin CPUs, isolate noisy neighbors, and feed accelerators without jitter. Work on raising the default kernel timer frequency from 250 Hz to 1000 Hz is already showing measurable boosts in Large Language Model (LLM) acceleration with negligible power cost.

While not a Linux default setting, some distros, like the Ubuntu low-latency kernels, now come with this as a standard setting.

Direct paths: cutting out the CPU middleman

Modern kernels allow GPUs to access memory, storage, and even peer devices directly, using technologies such as Nvidia's GPUDirect and peer-to-peer DMA. Combined with Compute Express Link (CXL) and improved Input/Output Memory Management Unit (IOMMU) handling, it enables accelerators to bypass the CPU when moving data. This eliminates bottlenecks that previously stalled ML training runs. This invisible plumbing is why AI clusters can scale out without collapsing under their own I/O.

What all this adds up to is that, when executives talk about "AI strategy," what they're not saying is that the unglamorous reality is that AI strategy depends on managing Linux at scale. It's all about patching kernels, hardening containers, and securing opaque workloads. AI may get the headlines, but Linux remains the operating system doing the actual work.

source

First-gen could debut with Windows 11 26H1



Nvidia hasn’t given up on its plans to ship a Windows on Arm chip this year. Supply-chain sources claim NVIDIA is still planning a Windows on Arm chip, with N1X-based laptops expected as early as Q1 2026. It will likely come with Windows 11 26H1 out of the box, as Windows Latest previously reported that this particular OS release is for new Silicon.



A Digitimes (Chinese) report based on supply-chain chatter says NVIDIA is trying to expand beyond GPUs and push deeper into PCs, especially Windows on Arm (WoA) laptops. The same roadmap also talks about newer chips after N1/N1X, moving toward N2 and N2X in 2027.

NVIDIA has not confirmed these timelines, and it usually avoids commenting on supply-chain leaks.

What are N1 and N1X Nvidia “AI PC” chips?

While N1 is likely for desktops, N1X is for notebooks, and Nvidia reportedly plans to announce N1X-based laptops running Windows 11 26H1 in Q1 (by the end of March 2026). Initially, Nvidia is focusing on consumer models, but there are plans for other variants, and we could see those PCs in Q2 2026.



N1X isn’t just a rumor, because NVIDIA is already using it in DGX Spark. The report says DGX Spark is based on N1X and includes the GB10 “superchip” with 128GB unified memory. Supply-chain sources also claim the same N1X platform will appear in Windows on Arm laptops as early as Q1 2026, including consumer models.

Multiple OEMs (Acer, Asus, Dell, Gigabyte, HP, Lenovo, MSI) are reportedly building their own DGX Spark systems.

According to Digitimes, Nvidia’s N1X notebooks were supposed to debut in late 2025, but they were pushed to 2026 due to Microsoft OS timing. It appears that Digitimes is referring to recent platform changes, which will begin shipping with Windows 11 26H1 in the coming months.



For those unaware, Microsoft officially confirmed that Windows 11 26H1 is for new Silicon, but it never said whether it’s only for Snapdragon X2 PCs.

“26H1 is not a feature update for version 25H2 and only includes platform changes to support specific silicon,” Microsoft noted in a blog post published in November 2025. 26H1 does not have exclusive features, but it’s based on a new platform release, which means it could include N1X and Snapdragon X2-related optimizations.

The report also claims that N1X chips were delayed due to weaker or uncertain notebook demand, Memory supply and pricing problems, which matter a lot for unified-memory designs.

Either way, it looks almost certain we’ll see Nvidia’s first Arm-based laptops in 2026.

More importantly, Nvidia is also working on N2 and N2X, which would be the next generation, with products launching starting around Q3 2027.

source

If you are a long-time Neowin reader, chances are you may be amongst those Windows users who had noticed a curious behavior: holding the Shift key while restarting doesn’t trigger a full cold reboot; instead, the system would do something slightly different.

For those not familiar, when a user held down the Shift key while restarting Windows 95, the system behaved differently than during a full cold reboot. Instead of cycling the hardware completely, Windows displayed “Windows is restarting” and attempted what was essentially a fast-restart. In a way, this was kind of like Fast Startup, which Microsoft introduced much later in Windows 8. If you attempt a Shift + Restart on Windows 11 and 10 you get into Windows Recovery Environment (WinRE).

Veteran Microsoft Windows developer Raymond Chen has explained how this worked. Chen, in his newly penned article on his The Old New Thing column, notes that this behavior was part of the old 16‑bit ExitWindows function when it received the EW_RESTARTWINDOWS flag.

If you are wondering, the ExitWindows function is a legacy function used to log off the interactive Windows user, while the EW_RESTARTWINDOWS parameter, as the name suggests, is used to restart the system.

Chen has explained that the shut down sequence began first with the 16‑bit Windows kernel itself, followed by the 32‑bit virtual memory manager, and then the CPU dropping back into real mode.

After this, control returned to the bootstrap program "win.com" with a special signal “Can you start protected mode Windows again for me?” thus instructing it to relaunch protected‑mode Windows. Hence, the code in win.com would then display the “Please wait while Windows restarts…” message as it tried to get the system back up as requested.

If you are trying to make sense of it, Win.com was essentially the executable file used to load different Windows versions based on DOS, like Windows 95. Meanwhile, Real mode Windows is an early design meant to run on PCs with minimal resources, like 192 KB of RAM and floppy drives, and Protected mode Windows is like the full OS version, with memory protection, GUI, and all.

Chen notes that by its nature of design, .com files claimed all conventional memory at launch, but in the case of win.com, it would release unused space to create one large contiguous block for protected‑mode Windows. So if another program had fragmented that memory space, the fast restart could not succeed, and win.com fell back to a full reboot. Otherwise, the fast restart continued as it re‑created the virtual machine manager and launched the graphical user interface (GUI), giving the user the impression of a seamless fast restart.

However, the process was not flawless as Raymond Chen adds, since some users reported that attempting two fast restarts in succession would lead to crashes, while others seemingly managed multiple fast restarts without issue. The likely explanation was that certain device drivers failed to reset properly, leaving corrupted memory that only revealed itself during shutdown.

source

I open Gmail dozens of times a day. It was where messages landed, and occasionally got starred or archived. I never expected it to be useful beyond email.

However, tucked away in the sidebar is a note-taking app I’d overlooked, even though I use Gmail every single day.

After I started using Google Keep directly inside Gmail, it changed how I handle notes, ideas, and tasks during the workday.

Instead of context-switching or letting thoughts slip away, I now capture them alongside my messages.

The Gmail sidebar I kept ignoring

If you use Gmail on desktop, you’ve probably noticed the slim sidebar on the right. It holds Google Calendar, Google Tasks, Google Keep, and Contacts.

I’d always dismissed it as clutter, thinking of it as something Google added to push its ecosystem.

Out of curiosity (and mild frustration with my inbox), I clicked the Keep icon one day. A familiar panel slid open, showing my notes exactly as they appear on my phone.

By keeping tools like Google Keep just a click away, Gmail offered a lightweight way to capture notes without switching tabs.

How I use Google Keep alongside Gmail

Using Google Keep in Gmail serves as a holding space for my thoughts and ideas while I manage my inbox.

If an email triggers an idea, a follow-up question, or something I want to think about later, I jot it down in Keep instead of leaving the message unread or starred.

That keeps my inbox focused on communication.

I also use Keep to extract context out of long email threads.

I’ll copy key details, such as dates, decisions, or action items, into a note so I don’t have to re-scan the entire conversation later.

It is beneficial for ongoing projects, where information often gets buried across multiple replies.

One limitation is that you can’t directly move an email into Google Keep. There’s no “send to Keep” button, which initially felt like a missed opportunity.

My workaround is simple: I open the email, copy its URL, and paste that link into a Keep note. That way, I can jump straight back to the original message whenever I need context.

This approach works well for ongoing threads. I’ll summarize the key points of the email in my own words, drop the Gmail link underneath, and move on.

It keeps my notes clean while still giving me a direct path back to the entire conversation.

Why this works better than copying emails elsewhere

Before I started using Keep inside Gmail, my default move was to copy email content into another app, whether it’s a note-taking app, a task manager, or a document.

However, every extra step made it less likely I’d capture the information in the first place.

Using Keep directly in Gmail is convenient. I don’t have to decide where something belongs or interrupt my flow by switching apps.

If a thought comes up while I’m reading an email, I capture it immediately.

There’s also less duplication and cleanup involved.

When I copy emails into other tools, I tend to over-save by copying entire threads or unnecessary details. With Keep, I’m more selective.

That makes the notes easier to scan and far more useful later.

Throughout the day, Keep becomes a lightweight scratchpad.

Drafting a reply, outlining a meeting agenda, or capturing something I want to revisit later all happen there, without ever leaving Gmail.

And because it syncs automatically, those notes are automatically updated on the mobile app.

The limitations of using Google Keep inside Gmail

As convenient as this setup is, it isn’t perfect.

Google Keep in Gmail is intentionally lightweight, and that simplicity means it won’t suit every workflow or every type of note.

The most significant limitation is structure. Keep doesn’t offer folders, nested notes, or advanced organization tools.

When your notes start piling up, relying on labels and search can feel a bit limiting compared to other note-taking apps.

It’s great for quick capture, but it’s not designed for managing large, long-term projects.

There’s also the lack of deeper email integration.

Since you can’t directly attach or move an email into Keep, referencing messages requires manual steps, such as copying links or summarizing content yourself.

Finally, the sidebar itself can feel cramped.

Writing longer notes or thinking through complex ideas in a narrow panel isn’t always comfortable.

When a note starts growing beyond a few lines, I often open Keep in a new tab.

Despite these limitations, I’ve stuck with this system because I’m not asking it to do too much.

Google Keep in Gmail doesn’t replace a full note-taking app; it’s there to catch thoughts before they disappear.

The Gmail feature that earned a permanent spot in my workflow

What surprised me most about using Google Keep inside Gmail is how little effort it takes. There’s no setup process or system to maintain.

By keeping note-taking right beside my inbox, I no longer have to trust my memory or let emails pile up as reminders.

It’s worth noting that this setup won’t replace a dedicated note-taking app. For me, its value lies in how seamlessly it fits into the day.

What once felt like a useless sidebar has transformed into one of the most consistently handy parts of Gmail, and now, I can’t imagine working without it.



source

FlyOOBE has received another feature update to give you more control over various parts of Windows 11. This tool is among the most popular third-party utilities for Windows debloating, and as the dev says in the latest release notes, it recently reached nearly 2.5 million downloads on GitHub, showing how much people want to clean Windows 11 from unnecessary stuff. The newest release, version 2.4, makes FlyOOBE even better by improving its capabilities to detect and remove AI components (the so-called Slopilot).

The release notes for FlyOOBE 2.4 state that the main goal of the update is to give users a choice. AI is not necessarily bad, but users should have the option to turn it off. If you do not want any of it on your PC, FlyOOBE is now better at detecting AI features and offers a deep cleaning feature via the RemoveWindowsAI script, which we reported earlier.

“We need to get beyond the arguments of slop vs sophistication…” - Satya Nadella
Agreed, but users still deserve a choice!!! The AI OOBE control has been refined and is now officially called Slopilot. Slopilot isnt anti-AI - it's pro user choice

In addition to that, there are several improvements for browser detection, extension engine, themes, and more. Here is the full changelog:

   • This update improves detection of AI-related (Slopilot) features across Windows 11 and adds optional deep cleanup capabilities via external tooling such as RemoveWindowsAI.

   • Users can now better understand, review, and disable AI components they don't want, transparently and on their own terms.

   • Improved browser detection in the Browser OOBE.

   • Updated Extensions engine for better coverage and accuracy.

   • Global search functionality expanded and refined across all sections.

   • Improved theme detection in the Personalization OOBE.

   • Minor core optimizations and internal cleanups.

You can download FlyOOBE 2.4 from its official GitHub repository, which is the only right place to get the app. You can find the link to it on Neowin's Software page as well.

source