She has put together a combination of kernel patches and userspace utilities that stop background apps from stealing VRAM away from whatever game you're playing.

The underlying issue is that when VRAM runs out, the kernel driver has no way to tell which memory matters more. A game and a browser tab look identical from the driver's perspective, so when something has to give, game memory often takes the hit.

It then ends up in GTT, a chunk of system RAM that the GPU can access, but over the PCIe bus rather than directly.

The fix is built on the dmem cgroup controller that she co-developed with Maarten Lankhorst from Intel and Maxime Ripard from Red Hat. It is already in the mainline Linux kernel, and it lets the driver treat foreground apps as higher priority when handing out VRAM.

That alone was not enough, though. Natalie has also written six kernel patches to fix a specific gap where VRAM pressure would cause new memory allocations to skip those protections entirely and end up in GTT anyway.

Two userspace utilities handle the rest: dmemcg-booster sets up the groundwork so the kernel protections actually activate, and a fork of KDE Plasma's Foreground Booster keeps track of which app is in the foreground so it gets first dibs on VRAM.

What this means for Linux gamers

Instead of performance slowly degrading over a session, games should now hold steady for as long as their own VRAM usage stays within budget. Natalie notes that most modern titles tend to stay within 8GB, so owners of 8GB GPUs should be in a much better spot with today's games.

While this applies to any GPU running the amdgpu driver, Intel GPUs on the xe driver have the necessary kernel support too, though real-world testing there is still pending.

Additionally, the developer has submitted a patch for nouveau, the open source NVIDIA driver.

How to get it

The six kernel patches are not in the mainline kernel, so getting them requires some extra steps depending on your setup. CachyOS users on Linux 7.0rc7-2 or later are already covered.

On other Arch-based distros, both utilities are in the AUR. For the kernel side, you can either pull the CachyOS kernel package from the repository or install linux-dmemcg from the AUR, which compiles Natalie's development branch.

The six patch files are also linked directly in the announcement blog for anyone who wants to apply them to a custom kernel build.

For those not on an Arch-based system, the realistic options are applying the patches manually to a self-compiled kernel or waiting for your distro to pick them up. Natalie has said her post will be updated if and when the work gets packaged by other distributions.

Suggested Read 📖: The Linux 7.0 Release is Here!

Late last year, at the 2025 Maintainers Summit, Sasha Levin pushed for some documented consensus, and what came out of it was human accountability for patches being non-negotiable, purely machine-generated submissions not being welcome, and tool use being disclosed.

He promised to put something in writing without committing to enforce it, and that work has now shipped with Linux 7.0.

What is it?

The new document is called AI Coding Assistants and lives in the kernel's process docs alongside the rest of the contribution guidelines. The short version is that AI-assisted contributions still need to comply with GPL-2.0-only; AI agents cannot add Signed-off-by tags; and patches that had AI help should carry an "Assisted-by" tag.

The Developer Certificate of Origin (DCO) is an important aspect that exists, so there is a human accountable for every patch. AI assistance does not change that hard requirement.

Basically, the human submitter reviews everything the AI produced, confirms it meets licensing requirements, and puts their own name on it with an appropriate mention that AI was used.

The Assisted-by tag format is Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]; for scenarios where either single or multiple tools were used. The document gives Assisted-by: Claude:claude-3-opus coccinelle sparse as an example.

Back then, Linus was not even convinced a dedicated tag was necessary and suggested the changelog body would do the job. But now, the kernel community seems to have settled on the tag anyway.

It's already in use

We covered this earlier in the week, but Greg Kroah-Hartman (GKH) seems to have had AI-assisted fuzzing running in his kernel tree for a while now, in a branch called "clanker." He started with the ksmbd and SMB code, found some potential issues, and submitted fixes with a note telling reviewers to verify everything independently before trusting any of it.

That is just about the workflow the new policy was written around. AI surfaces issues, a human with decades of kernel experience decides what is real, writes the fix, and takes responsibility. GKH being the one doing it is not a surprise given he is the stable kernel maintainer and has probably dealt with more bad patches than the others.

Other projects have gone in a different direction. Gentoo banned AI-generated contributions entirely in 2024, with its council citing copyright risk, code quality, and ethical concerns.

NetBSD's commit guidelines put LLM-generated code in the "tainted code" category, requiring written approval from the core developers before any of it goes in.

In contrast, Linux is not banning anything. Whether that turns out to be the sensible call or just a lenient one will depend on how seriously people actually take the "a human reviewed this" part.

Suggested Read 📖: Is a Clanker Being Used in Linux Development?

But that's a story for another time. For now, let's focus on what Linux 7.0 brings to the table.

Head penguin, Linus Torvalds, had the following words to say regarding the release:

The last week of the release continued the same "lots of small fixes" trend, but it all really does seem pretty benign, so I've tagged the final 7.0 and pushed it out.

I suspect it's a lot of AI tool use that will keep finding corner cases for us for a while, so this may be the "new normal" at least for a while. Only time will tell.

Linux Kernel 7.0: What's New?

The release is here, and before getting into the improvements, there is one thing worth getting out of the way first.

This is not a long-term support release. If your priority is stability and extended maintenance, this is not the kernel to land on. Instead, you could opt for Linux kernel 6.18, which is supported until December 2028.

Intel Upgrades

Linux 6.19 already added audio support for Intel Nova Lake S, but the standard Nova Lake (NVL) variant was left out. That's fixed in 7.0, and the difference between the two in terms of specs is mainly in core count (4 vs. 2).

Intel Arc users get something useful too. The Xe driver now exposes a lot more temperature data through the HWMON interface. Previously you got a single GPU core reading; now you get shutdown, critical, and max temperature limits, plus memory controller, PCIe, and individual vRAM channel temperatures.

Also for Panther Lake, GSC firmware loading and Protected Xe Path (PXP) support are in.

And lastly, Diamond Rapids (the upcoming Xeon successor to Granite Rapids) gets NTB driver support, which handles high-speed data transfers between separate systems over PCIe. It is expected to be helpful for distributed storage and cluster setups.

AMD Refinements

While the Zen 6 series of CPUs are still a while out, the kernel is already getting ready for it. Linux 7.0 merges perf events and metrics support for AMD Zen 6, covering performance counters for branch prediction, L1 and L2 cache activity, TLB activity, and uncore events like UMC command activity.

All of that is mainly useful for developers and admins doing performance profiling ahead of launch, and not something the average user will notice.

For virtualization, KVM picks up support for AMD ERAPS (Enhanced Return Address Predictor Security), a Zen 5 security feature. In VM scenarios, this bumps the Return Stack Buffer from 32 to 64 entries, letting guests make full use of the larger RSB.

AMD is also laying the groundwork for next-gen GPU hardware in 7.0, enabling new graphics IP blocks for what looks like an upcoming RDNA 4 successor and another RDNA 3.5 variant.

There are also hints of deeper NPU integration with future Radeon hardware, but AMD hasn't announced anything yet, so exact product details remain a mystery for now.

Better Storage Handling

XFS gets one of the more interesting additions this release called autonomous self-healing. A new xfs_healer daemon, managed by systemd, watches for metadata failures and I/O errors in real time and triggers repairs automatically while the filesystem stays mounted.

Btrfs picks up direct I/O support for block sizes larger than the kernel page size, falling back to buffered I/O when the data profile has duplication. There's also an experimental remap-tree feature, which introduces a translation layer for logical block addresses that lets the filesystem handle relocations and copy-on-write operations without physically moving or rewriting blocks.

EXT4 sees better write performance for concurrent direct I/O writes to multiple files by deferring the splitting of unwritten extents to I/O completion. It also avoids unnecessary cache invalidation and forced ordered writes when appending with delayed allocation.

Miscellaneous Changes

Wrapping up this section, we have some other notable changes that made it into this release:

• RISC-V gains user-space control-flow integrity (CFI) support.
• WiFi 8 Ultra-High Reliability (UHR) groundwork lands in the networking stack.
• Security bug report documentation gets an overhaul to help AI tools send more actionable reports.
• Rust support is officially no longer experimental, with the kernel team formally declaring it is here to stay.
• ASUS motherboards, including the Pro WS TRX50-SAGE WIFI A and ROG MAXIMUS X HERO, now have working sensor support.

Installing Linux Kernel 7.0

As always, those on rolling distros like Arch Linux and other distros like Fedora and its derivatives will get this new release very soon. For others on distros like Debian, Linux Mint, Ubuntu, MX Linux, etc. You will most likely not receive this upgrade.

If that doesn't work for you, then you could always install the latest mainline Linux kernel on your Ubuntu setup. And, this goes without saying, this is risky. If you end up borking your system, we are not to blame for it.

In this beginner's guide, we will look at the built-in screenshot tool in Linux Mint and the keyboard shortcuts you can use right away.

The GUI screenshot tool that you don't want to miss

Linux Mint provides a simple graphical interface for those who prefer a GUI solution for taking screenshots.

Beyond the basic options, the tool also includes a few useful features. Let’s take a look at them next.

First, open the Screenshot tool by searching for it in the start menu.

The interface is simple and easy to understand. There are three main options:

• Capture Screen: Takes a screenshot of the entire screen
• Capture Window: Captures the active window
• Capture Selection: Lets you select a specific area using left-click and drag to capture.

After choosing the method, click the Take Screenshot button at the top left of the window.

Show mouse cursor in screenshot

In the Screenshot tool, you will find an option called Show Pointer. Enable this if you want the mouse pointer to be visible in your screenshots.

Take screenshot with a delay

You can also set a small delay before taking a screenshot.

In the Screenshot tool, enter a value in seconds under the Delay in Seconds option.

Once set, the tool will wait for the specified time before capturing the screenshot when using the GUI. For example, if you set it to 5 seconds, the screenshot will be taken after a 5 second delay.

Using keyboard shortcuts

If you prefer not to open a GUI app every time you take a screenshot, that is not a problem. Linux Mint provides keyboard shortcuts that let you quickly capture the screen in different ways.

Take the screenshot of entire screen

You can press the PrtScr key on your keyboard to capture the entire screen.

After taking the screenshot, you will be prompted to either save it with a name or copy it to the clipboard. This works well for basic use.

However, this can feel limited if you only want to capture a small part of the screen. The good news is that Linux Mint also provides an easy way to do that.

Take the screenshot of an area

To take the screenshot of a specific area, use the Shift + PrtScr shortcut.

Your screen will dim slightly and the cursor will change to a plus sign. Click, hold, and drag to select the area you want to capture.

Once you release the mouse button, you can choose to copy the screenshot or save it.

Take screenshot of a window

Sometimes, you may want to capture only the currently active window. While you can do this using the area selection method, using a shortcut is much more convenient.

Press Alt + PrtScr to take a screenshot of the active window.

There are a few things to keep in mind. If a menu is open inside the window, like a top menu or a right-click context menu, this shortcut may not work.

Also, if a dialog box is open, the tool will capture whichever window is active at that moment, whether it is the main window or the dialog.

Record the screen

Many people do not realize that Linux Mint also includes a built-in screen recorder. It is not visible in the menus, so it is easy to miss.

Press Shift + Ctrl + Alt + R to start recording your screen. Use the same key when a recording is active to stop recording.

This is a basic tool, so do not expect features like those in dedicated applications such as OBS Studio or SimpleScreenRecorder. It simply records your entire screen.

When you stop the recording, the video file is saved in the Videos folder inside your Home directory.

Custom Shortcuts

In the previous section, we saw that the GUI tool offers options like delay and showing the mouse pointer, which are not available with the default keyboard shortcuts.

However, this does not mean you are limited. In Linux Mint, you can create custom shortcuts to include these actions as well.

The screenshot options

Before setting up custom screenshot shortcuts, it helps to understand the available options. Linux Mint uses the GNOME Screenshot tool for both the GUI and keyboard based screenshots.

GNOME Screenshot provides several useful options, along with many more that you can explore in its man page.

• gnome-screenshot -w: Take the screenshot of current active window.
• gnome-screenshot -a: Take the screenshot of a select region by click and drag.
• gnome-screenshot -d 5: Add a 5 second delay before taking a screenshot of the entire screen.
• gnome-screenshot -d 5 -p: Apply a 5 second delay and include pointer in the screenshot.
• gnome-screenshot -d 5 -a, gnome-screenshot -d 5 -w: Take screenshot of select area/window respectively with a 5 second delay.

Setting custom screenshot shortcuts

Search for and open Keyboard from the start menu.

Go to the Shortcuts tab and then select Custom Shortcuts. Click on the Add custom shortcut button.

Now, enter a name for the shortcut. For example, you can use "Take screenshot of an area with a delay" in the Name field.

In the command field, enter the required command. For example, use gnome-screenshot -d 5 -a, and then click the Add button.

The command will now be listed. To assign a shortcut, select it under Keyboard shortcuts and click on the Unassigned option in the Keyboard bindings section.

You will be prompted to press a key combination. Press the shortcut you want to use.

You can repeat the same steps to create and assign shortcuts for other commands based on your needs.

Other screenshot tools

Sometimes, basic screenshots are not enough. You may want to annotate an image or add borders and other adjustments.

These are image editing features, and they are not available in the default Screenshot tool in Linux Mint.

For such needs, you can use third party screenshot tools that offer more control and customization.

We have a separate article that covers screenshot tools you can use in more detail. You can refer to it to find options that suit different needs and use cases.

As a quick note, Flameshot and Ksnip are two good screenshot tools you can use for editing and customization. You can also use Gradia that also provides basic editing.

Did you find it useful? Feel free to share your thoughts in the comments.

Messages are routed through an onion network rather than a central server, and the combination of no-metadata messaging, anonymous sign-up, and decentralized architecture has earned it a loyal following among privacy-conscious users.

Unfortunately, the project has sent out a mayday call as it risks closure.

A call for help

The Session Technology Foundation (STF) sent out what can only be described as a distress signal, announcing that the app's survival is now in serious peril. The day it was posted on was also the last working day for all paid staff and developers at the STF.

From that point on, Session is being kept running entirely by volunteers.

The donations that they received earlier are enough to keep critical infrastructure online until July 8, but not nearly enough to retain a development team. With nobody left on payroll, development has been paused.

Due to that, introducing new features is off the table, existing bugs will most likely go unaddressed, and the STF says new releases are unlikely during this period.

Session co-founder Chris McCabe had already flagged the trouble coming. In a personal appeal published earlier in March, he wrote that the organizations safeguarding Session had faced many challenges over the years and that the project's very survival was now at risk.

He had concluded by appealing that:

The project is on a path to self-sustainability, but the future is fragile. If every Session user contributed just one dollar, it would go a long way towards Session reaching sustainability. If you've ever considered donating, now is the time to act.

The above didn't accomplish enough to change the outcome, so the Session folks had to sound the alarm. The foundation says it needs $1 million to complete the work still in progress.

That includes Protocol v2, which adds forward secrecy (PFS), post-quantum cryptography, and improved device management, as well as Session Pro, a subscription tier intended to put the project on a self-sustaining footing.

If that goal is hit, the STF says it hopes Session could stand on its own without needing to go back to the community for more.

As of writing, $65,000 of that $1 million has been raised. Anyone who wants to see this privacy-focused messaging app survive, especially at a time when surveillance is only getting worse, can donate at getsession.org/donate.

Suggested Read 📖: Session's Other Co-Founder Thinks You Don't Need to Ditch WhatsApp

The Linux switch is not the only move on the table. France's national health insurance body, CNAM, is migrating 80,000 of its agents to a set of homegrown tools: Tchap for messaging, Visio for video calls (more on this later), and France transfert for file transfers.

The country's national health data platform is also set to move to a sovereign solution by the end of 2026.

Beyond the immediate moves, the seminar laid out a broader plan. DINUM will coordinate an interministerial effort built around forming coalitions between ministries, public operators, and private sector players, with interoperability standards at the core (the Open Interop and Open Buro initiatives are specifically named).

Every French ministry, including public operators, will be required to submit its own non-European software reduction plan by Autumn 2026.

The plan is expected to cover things like workstations, collaboration tools, antivirus, AI, databases, virtualization, and network equipment. A first set of "Industrial Digital Meetings" is planned for June 2026, where public-private coalitions are expected to be formalized.

Speaking on this initiative, Anne Le Hénanff, Minister Delegate for Artificial Intelligence and Digital Affairs, added that (translated from French):

Digital sovereignty is not optional — it is a strategic necessity. Europe must equip itself with the means to match its ambitions, and France is leading by example by accelerating the shift to sovereign, interoperable, and sustainable solutions.

By reducing our dependence on non-European solutions, the State sends a clear message: that of a public authority taking back control of its technological choices in service of its digital sovereignty.

You might remember, a few months earlier, France set out on a similar path for video conferencing. The country mandated that every government department switch to Visio, its homegrown, MIT-licensed alternative to Teams and Zoom by 2027.

Part of the broader La Suite Numérique initiative, it had already been tested with 40,000 users across departments before the mandate was announced. So this move looks like an even more promising one, and we shall keep an eye on how this pans out.

Suggested Read 📖: ONLYOFFICE Gets Forked

Now that terminology has made its way into the Linux kernel thanks to Greg Kroah-Hartman (GKH), the Linux stable kernel maintainer and the closest thing the project has to a second-in-command.

He has been quietly running what looks like an AI-assisted fuzzing tool on the kernel that lives in a branch called "clanker" on his working kernel tree. Before you ask, fuzzing is a method of automated software testing that bombards code with unexpected, malformed, or random inputs to trigger crashes, memory errors, and other misbehavior.

It is a critical line of defense for a massive codebase like Linux.

How it started

It began with the ksmbd and SMB code. GKH filed a three-patch series after running his new tooling against it, describing the motivation quite simply. He picked that code because it was easy to set up and test locally with virtual machines.

What the fuzzer flagged were potential problems specific to scenarios involving an "untrusted" client. The three fixes that came out of it addressed an EaNameLength validation gap in smb2_get_ea(), a missing bounds check that required three sub-authorities before reading sub_auth[2], and a mechToken memory leak that occurred when SPNEGO decode fails after token allocation.

GKH was very direct about the nature of the patches, telling reviewers: "please don't trust them at all and verify that I'm not just making this all up before accepting them."

It does not stop there. The clanker branch has since accumulated patches across a wide range of subsystems, including USB, HID, WiFi, LoongArch, networking, and more.

Who is GKH?

If you are not well versed with the kernel world, GKH is one of the most influential people in Linux development.

He has been maintaining the stable kernel branch for quite a while now, which means every long-term support kernel that powers servers, smartphones, embedded devices, and pretty much everything else running Linux passes through his hands.

He also wrote Linux Kernel in a Nutshell back in 2006, which is freely available under a Creative Commons license. It remains one of the more approachable references for anyone trying to understand kernel configuration and building, and it is long overdue for a new edition (hint hint).

Linus has been thinking about this too

Speaking at Open Source Summit Japan last year, Linus Torvalds said the upcoming Linux Kernel Maintainer Summit will address "expanding our tooling and our policies when it comes to using AI for tooling."

He also mentioned running an internal AI experiment where the tool reviewed a merge he had objected to. The AI not only agreed with his objections but found additional issues to fix.

Linus called that a good sign, while asserting that he is "much less interested in AI for writing code" and more interested in AI as a tool for maintenance, patch checking, and code review.

AI should assist, not replace

There is an important distinction worth making here. What GKH appears to be doing here is not having AI write kernel code. The fuzzer surfaces potential bugs; a human with decades of kernel experience reviews them, writes the actual fixes, and takes responsibility for what gets submitted.

If that's the case, then this is the sensible approach, and it mirrors what other open source projects have been formalizing. LLVM, for instance, adopted a "human in the loop" AI policy earlier this year, requiring contributors to review and understand everything they submit, regardless of how it was created.

Suggested Read 📖: Greg Kroah-Hartman Bestowed With The European Open Source Award

But it is also the same company that bakes telemetry into Windows by default and has been aggressively pushing Copilot AI into every corner of its software. That last part especially has been nudging a growing number of people toward open alternatives.

And now, a wave of developer account suspensions has given some open source developers a new headache.

What's happening?

Microsoft rolled out mandatory account verification for all partners enrolled in the Windows Hardware Program who had not completed verification since April 2024. The requirement kicked in on October 16, 2025, giving partners 30 days from notification to verify their identity with a government-issued ID.

Plus, that ID has to match the name of the Partner Center primary contact. Miss the deadline or fail verification, and your account gets suspended with no further submissions allowed.

This matters because signing Windows kernel drivers requires one of these accounts. Without it, developers cannot push driver-signed updates for Windows, and Windows will flag unsigned drivers, blocking them from loading at the kernel level.

Three major open source projects found this out the hard way. VeraCrypt, WireGuard, and Windscribe all had their developer accounts suspended, cutting off their ability to ship updates on Windows.

VeraCrypt developer Mounir Idrassi was the first to go public. In a SourceForge forum post, he wrote that Microsoft had terminated his account with no prior warning, no explanation, and no option to appeal.

Repeated attempts to reach Microsoft through official channels got him nothing but automated replies. The suspension hit his day job too, not just VeraCrypt.

WireGuard creator Jason Donenfeld hit the same wall a couple of weeks later, when he went to certify a new WireGuard kernel driver for Windows and found his account showing as access restricted. He eventually tracked down a Microsoft appeals process, but it carried a 60-day response window.

Windscribe's situation was arguably the messiest. The company says it had held a verified Partner Center account for over eight years and spent more than a month trying to sort things out before going public.

Moreover, once an account is suspended, Partner Center blocks users from opening a support ticket directly.

What now?

This eventually got Microsoft's attention as Scott Hanselman, VP and Member of Technical Staff at Microsoft and GitHub stepped in on X to say the accounts would be fixed. He pointed to the October 2025 blog post (linked earlier) and said the company had been sending emails to affected partners since then.

Scott confirmed he had personally reached out to both Mounir and Jason to get their accounts unblocked, and that fixes were already in progress.

Anyway, this doesn't look good, and leaving developers of critical security software without recourse for weeks only erodes trust. But, in the end, this won't really affect a behemoth like Microsoft, who has a dominating hold on the operating system market.

Suggested Read 📖: Proton Workspace and Meet launched as alternatives to Big Tech offerings

Git, of course, is a version control system primarily used by programmers.

But Theena makes a strong case that Git and plain text are the best tools a writer can use. Not just for backup but for building a writing practice that is truly their own..

At its core, the argument is about breaking free from platform dependency, long-term preservation, and treating your body of work as something worth designing around rather than just storing somewhere convenient.

Here are other highlights of this edition of FOSS Weekly:

• sudo tips and tweaks.
• Apt's new version has useful features.
• Opera GX arriving as a gaming browser for Linux.
• A Linux driver proposal to catch malicious USB devices.
• And other Linux news, tips, and, of course, memes!

📰 Linux and Open Source News

Not open source software but Opera GX, the gaming-focused Chromium browser that's been on Windows and macOS for years, has finally landed on Linux. Sourav took the early access build for a spin and tested the features it's known for, like GX Control for capping RAM and CPU usage while gaming and GX Cleaner for cleaning up junk data.

The Linux kernel is finally dropping i486 support, queued for Linux 7.1. The first patch removes the relevant Kconfig build options, with a fuller cleanup covering 80 files and over 14,000 lines of legacy code still to follow.

Proton has launched two new things: Proton Workspace, a bundled suite of all their services aimed at businesses looking for a privacy-first alternative to Google Workspace or Microsoft 365, and Proton Meet, an end-to-end encrypted video conferencing tool using the open source MLS protocol.

A proposal has been submitted to the Linux kernel mailing list for a new HID driver called hid-omg-detect that passively monitors USB keyboard-like devices for suspicious behavior.

Another proposal, but for Fedora was recently struck down. It looked to move per-user environment variable management from shell RC files into systemd.

Remember the glass UI from the Windows 7 era? KDE is considering bringing back the older classic Oxygen and Air themes. These themes will be optional, of course.

Anthropic, the company behind Claude AI, has donated $1.5 million to Apache Software Foundation. The donation aims to secure the open source stack AI tools depend on.

🧠 What We’re Thinking About

Firefox has been losing ground for a decade, and Mozilla is trying something new. A built-in VPN and a growing set of AI features. Roland's piece looks at whether either of those things is likely to actually work.

Puter, the open source browser-based desktop OS, has added ONLYOFFICE to its app marketplace, giving it a full office suite covering documents, spreadsheets, presentations, and PDF editing.

🧮 Linux Tips, Tutorials, and Learnings

Not many people know that sudo command's behavior can be tweaked as well. Here are a few sudo tweaks.

Tennis is a Zig-written terminal tool that renders CSV files as clean, color-coded tables with solid borders and auto-detected themes.

APT package manager's latest version 3.2 has a rollback feature. Sourav briefly tested it.

📚 Linux eBook bundle (don't miss)

No Starch Press needs no introduction. They have published some of the best books on Linux. And they are running an ebook bundle deal on Humble Bundle.

I highly recommend checking it out and getting the bundle.

Plus, part of your purchase supports Electronic Frontier Foundation (EFF).

👷 AI, Homelab and Hardware Corner

The Linux kernel dropped i486 support and added GD-ROM driver support for the Sega Dreamcast in the same breath.

✨ Apps and Projects Highlights

Hideout is a minimal GTK4/Adwaita desktop app for file encryption and decryption, powered by GnuPG.

📽️ Videos for You

Here are some Linux terminal tricks to save you time.

💡 Quick Handy Tip

You can copy a file in Nautilus by pressing Ctrl+C, then press Ctrl+M to paste it as a symbolic link instead of an actual copy. This is a handy way to create a symlink without ever needing to open a terminal!

🎋 Fun in the FOSSverse

In this members-only crossword, you will have to name systemd's ctl commands.

An appropriate meme on the OS-level age verification topic.

🗓️ Tech Trivia: On April 8, 1991, a small team at Sun Microsystems quietly relocated to work in secret on a project codenamed "Oak", a programming language that would eventually be renamed Java and go on to become one of the most widely used languages in the world, powering everything from Android apps to enterprise software.

🧑‍🤝‍🧑 From the Community: A FOSSer is wondering if anyone has ever jailbroken a Kindle for KOReader use.

It has been around for decades, and if you are on a Debian-based distro, then you have almost certainly used it without giving it much thought. That said, it has seen active development in the last couple of years.

We covered the APT 3.0 release this time last year, which kicked off the 3.x series with a colorful new output format, the Solver3 dependency resolver, a switch from GnuTLS/GnuPG to OpenSSL, and Sequoia for cryptographic operations.

The 3.1.x cycle that followed has now closed out with APT 3.2 as the stable release, and it brings some notable changes with it.

What do you get with Apt 3.2?

The biggest additions with this release are transaction history with rollback support, some new commands, and per-repository package filtering.

APT now keeps a log of every package install, upgrade, and removal. You can view the full list with apt history-list, which shows all past operations with an ID assigned to each. To see exactly what packages were affected in a specific operation, you can use apt history-info <ID>.

From there, apt history-undo <ID> can be used to reverse a specific operation, reinstalling removed packages or removing installed ones as needed. If you undo something mistakenly and want it back, run apt history-redo <ID> to reapply it.

For cases where you want to revert everything back to the state at a particular point, apt history-rollback <ID> does that by undoing all operations that happened after the specified ID. Use this with care, as it makes a permanent change.

apt why and apt why-not are another set of new additions that let you trace the dependency chain behind a package. Run apt why <package> and APT will tell you exactly what pulled it onto your system. Run apt why-not <package> and it will tell you why it is not installed.

Similarly, Include and Exclude are two new options that let you limit which packages APT uses from a specific repository. Include restricts a repo to only the packages you specify, and Exclude removes specific packages from a repo entirely.

Solver3, which shipped as opt-in with APT 3.0, is now on by default. It also gains the ability to upgrade packages by source package, so all binaries from the same source are upgraded together.

Additionally, your system will no longer go to sleep while dpkg is running mid-install and JSONL performance counter logging is also in, though that is mostly useful for developers.

If all of that's got you interested, then you can try Apt 3.2 on a Debian Sid installation as I did below or wait for the Ubuntu 26.04 LTS release, which is reportedly shipping it.

How to use rollback on Apt?

After installing some new programs using APT, I tested a few commands to see how rollback and redoing transactions worked. First, I ran sudo apt history-list in the terminal and entered my password to authorize the command.

The output was a list of APT transactions that included the preparatory work I had done to switch to Debian Sid from Stable, as well as the two install commands to get Vim and Nala installed.

Next, I ran sudo apt history-info 4, the number being the ID of the transaction, and I was shown all the key details related to it, such as the start/end time, requested by which user, the command used, and packages changed.

After that, I ran sudo apt history-undo 4 to revert the Vim installation and sudo apt history-redo 4 to restore the installation; both of these commands worked as advertised.

Finally, I tested sudo apt history-rollback 3 to get rid of Nala, and the process was just about the same as before, with me being asked to confirm changes by typing "Y".

When I tried to run apt history-redo for this one, the execution failed as expected.

💬 Do these new additions look useful to you? Can't be bothered? Let me know below!

If you have used the internet today, you have almost certainly touched something the ASF maintains. Some of its projects like Kafka, Spark, Cassandra, the Apache HTTP Server, are not some niche tools, but a critical part of the modern IT infrastructure.

The ASF does not sell anything. It runs on donations, and without sustained funding, the infrastructure behind all of that software does not maintain itself.

Anthropic's framing for the donation is essentially that AI runs on this stuff and someone has to fund it. As AI development moves forward more quickly, the open source foundations underneath it need to be in good shape to keep up.

On the topic, Ruth Suehle, President of the Apache Software Foundation, added that:

Open source software is the foundation of modern digital life — largely in ways the average person is completely unaware of — and ASF projects are a critical part of that. When it works, nobody notices, and that’s exactly the goal.

But that kind of reliability isn’t a given. It is the result of sustained investment in neutral, community-governed infrastructure by each part of the ecosystem. Support like Anthropic’s helps ensure long-term strength, independence, and security of the systems that keep the world running.

Similarly, Vitaly Gudanets, Chief Information Security Officer at Anthropic, said that:

AI is accelerating rapidly, but it’s built on decades of open source infrastructure that must remain stable, secure, and independent. Supporting the Apache Software Foundation is a direct investment in the resilience and integrity of the systems that modern AI — and the broader software ecosystem — depend on.

Some thoughts

You might remember Anthropic was part of a similar donation campaign back in March, when the Linux Foundation announced $12.5 million in grants to strengthen open source software security. Anthropic was one of seven contributors to that pool, alongside AWS, Google, Google DeepMind, GitHub, Microsoft, and OpenAI.

That funding was managed by Alpha-Omega and the Open Source Security Foundation (OpenSSF), with the goal of helping open source maintainers deal with the growing flood of AI-generated vulnerability reports they simply do not have the bandwidth to handle.

It is great to see open source receiving monetary support, but the smaller players who are equally important in the ecosystem also need to be supported better. Big donations like this tend to flow toward well-established foundations, while the countless smaller projects that hold up just as much critical infrastructure quietly struggle for resources.

Both were announced at PyTorch Conference Europe in Paris, and the two now join DeepSpeed, Ray, vLLM, and PyTorch itself as foundation-hosted projects.

Moreover, the foundation has confirmed that ExecuTorch, Meta's solution for running PyTorch models on edge and on-device environments, is being merged into PyTorch Core.

If you were looking for the why, it is fairly straightforward.

Both moves come as AI teams increasingly focus on getting models into production rather than just training them. Running kernels efficiently across different hardware and keeping model files safe to load are two problems the ecosystem has been dealing with for a while.

Talking about Helion joining up, Matt White, the CTO of the PyTorch Foundation, added that:

Helion gives engineers a much more productive path to writing high-performance kernels, including autotuning across hundreds of candidate implementations for a single kernel.

As part of the PyTorch Foundation community, this project strengthens the foundation for an open AI stack that is more portable and significantly easier for the community to build on.

Luc Georges, Chief Open Source Officer, Hugging Face echoed similar excitement:

Safetensors joining the PyTorch Foundation is an important step towards using a safe serialization format everywhere by default. The new ecosystem and exposure the library will gain from this move will solidify its security guarantees and usability. Safetensors is a well-established project, adopted by the ecosystem at large, but we're still convinced we're at the very beginning of its lifecycle: the coming months will see significant growth, and we couldn't think of a better home for that next chapter than the PyTorch Foundation.

What does this mean?

The PyTorch Foundation is a Linux Foundation-hosted organization that acts as the vendor-neutral home for PyTorch and a growing set of open source AI projects. The main goal here is to keep governance and technical direction community-driven rather than tied to any single company's whims.

The Linux Foundation is the broader stewardship body behind over 1,000 open source projects, covering everything from the Linux kernel and Kubernetes to OpenSSF. The PyTorch Foundation sits under that umbrella, giving its projects access to LF's governance infrastructure and oversight.

Helion comes in as a tool that makes writing the low-level code that runs AI models on GPUs significantly less painful. It handles a lot of the tedious groundwork automatically, and finds the best configuration for your hardware on its own.

Whereas Safetensors is a file format for storing and sharing AI model weights that doesn't come with the security baggage of older formats.

Both themes trace their roots back to the KDE 4 era. Oxygen shipped as the default theme from KDE 4.0, defined by its dark tones and glassy aesthetic. It held that spot until KDE 4.3, when Air took over as the default, bringing a lighter look built around transparency and white as its base color.

While Oxygen stuck around into the Plasma 5 and 6 eras, it did so in an increasingly broken condition, and Air eventually got dropped from Plasma entirely.

Now, both are getting a second shot thanks to the restoration effort led by KDE contributor Filip Fila, alongside the original Oxygen designer Nuno Pinheiro and several other KDE developers.

On the Oxygen side, the panel has been fully reworked and is now orientation-aware, so vertical panels actually behave correctly. A minimized window indicator and a proper switch design were both missing entirely and have now been added.

Similarly, adaptive opacity is now supported and enabled by default, and the color scheme bug that was causing readability issues in widgets like System Monitor has been fixed.

Air needed its transparency restored to match its original KDE 4 character. That is done now, with blur added behind widgets, improving readability and visual appeal in the process. The panel has also been reworked, a new header and footer design has been added, and Air now has its own switch SVGs.

Why now after all this time? Well, KDE's 30th anniversary coincides with the Plasma 6.7 release, and the people behind this want to ship these historically significant themes for the occasion.

As of writing, 26 of 40 checklist items have been completed (linked below), with some pending work including gradient banding fixes in Oxygen, missing SVGs for checkmarks, radio buttons, toolbar, and menubar items across both themes, and a timer SVG for Air.

And if you want to see what that progress looks like, continue reading! 😬

How do these compare to Breeze?

I checked out how Plasma's default Breeze theme compared to Oxygen and Air on a KDE Neon setup, and I must say, things are looking promising. The themes have things like the panel styling, widget backgrounds, and the new switch designs in place.

I specifically took a look at the panel and widgets, and these looked very clean, feeling like they belonged in the modern Plasma experience, which is not something you would expect from themes this old.

One thing worth noting is that the icons stayed as Breeze regardless of which Plasma Style I picked.

As for the difference between these, Breeze is flat by design. Minimal, no frills, gets out of your way kind. Oxygen and Air are not like that, bringing visible depth and some bling to the desktop, but in different ways.

While Air leans hard into transparency, making panels and widgets look light and barely there, Oxygen goes the other direction with darker gradients and more visual weight across the board.

Personally, I prefer Oxygen as it looks a lot like Windows 7's Aero, which I quite liked back in the day.

You can try these out too!

First, you have to download the files for Oxygen and Air on a KDE Plasma-equipped system. Next, you have to go into System Settings > Appearance > Colors & Themes > Plasma Style.

Here, click on "Install from File..." and select a theme file to install it. Repeat for the other one, then select whichever theme you want and hit "Apply" on the bottom-right.

If you want to stay in sync with the development of these, you can keep an eye out on the GitLab issue tracker and Telegram group for this project.

It has everything you would expect in a desktop. A file system, app store, text editor, IDE, video player, etc. The video below shows it in action. See, how somooth it is.

Why not give it a try yourself here.

Puter has proper office tools now

As you can see in the video above, Puter has a text editor which is capable of taking notes. But it is not enough for rich text format and proper documents.

Now, they are offering ONLYOFFICE in their app center. The open source ONLYOFFICE is a collaborative productivity suite like MS Office 365.

With this integration, you also get spreadhseet, pdf editor and presentation tools. This make the online browser based desktop operating system more capable and complete than ever.

Like Nextcloud but with filesystem

If you ever used Nextcloud, you know that the cloud storage also has office integration with ONLYOFFICE and Collabora (sort of LibreOffice online). The difference I see here is that Puter gives you the entire filesystem, not just cloud storage and productivity tools. This is excellent for people who prefer the comfort of a desktop like environment in the cloud.

Experiencing the office experience in online OS

I have used ONLYOFFICE in Nextcloud in the past. I have noticed that it does take a bit of time to load the application in both Puter and Nextcloud. Sure, we cannot expect a full office app to open in a jiffy like a text editor. These applications take longer to start, even on local desktop system. It takes pretty much the same in Puter.

Creating the document was smooth. The in-built grammar checker worked well too. For some reason, the Save button on the top interface did not work. The print and open and other buttons were working fine and so did the keyboard shortcut. Not a dealbreaker but something worth mentioning. I am sure the Puter team will fix it in the next iteration.

Since I didn't have any complicated word document with macros and images, I uploaded a csv file with 1000 rows and opened it in ONLYOFFICE. I was expecting it to take some time to load and struggle but it did not. Obviously, that's a good thing.

I recorded it in a video so that you can see if yourself if it takes too long or lags while moving down the sheet.

I also created a simple presentation slide. Unsurprisingly again, it went smoothly. ONLYOFFICE actually works quite well in a web browser. That is their USP, afterall.

The entire ONLYOFFICE integration gives Puter the complete desktop OS features. After all, ONLYOFFICE also has PDF editor and diagram app (Visio alternative). All your office needs are sorted, that too in a web browser.

Many more apps at your disposal

Puter Marketplace (app store) has many more applications that you can explore and use as per your requirement. From games to coding to AI, there are plenty of options.

We had covered Puter last summer. In less than a year, Puter has grown and created an ecosystem that is admirable. I was skeptical of the actual, real-world usage of an online operating system that lives in a web browser but it seems there is demand for such a thing. Good to see an open source project filling up this demand.

If you are coming across Puter OS for the first time, you should know that the files you uploaded or created will be available as long as the browser session is not deleted. But if you want to save them for longer period, you can create an account with Puter. If you see a good use of Puter in your workflow or it could help your team (if you have one), you can also self-host it or pay to Puter for their managed hosting.

Opera is a name that never fully disappeared from my mind because it has been consistent in delivering a browsing experience that many people prefer. I used it for a while, mostly for the free built-in VPN, before eventually moving to Firefox when I felt it was time for a change.

However, they also have a gaming-focused browser called Opera GX, which has been available on Windows and macOS for some time now. Earlier this year, we got word that a Linux port was in the works, and it did eventually arrive.

Curious about what took them so long, I asked Maciej Kocemba, Product Director at Opera GX, and he had this to add:

Bringing Opera GX to Linux has been a priority for us for some time now, especially since we've seen such public support among the community. One group even launched an online petition that collected several hundred signatures, which was pretty cool to see.

With gaming on Linux growing so fast right now, this is the perfect moment for us to bring a browser designed for gamers to a platform that values customization and control as much as we do. We’re so happy to finally make it available to this community of users, and we're eager to see how they'll take advantage of the GX features they’ve been waiting for.

That got me hyped enough to see for myself what a gaming browser actually feels like and whether there was anything a regular browser couldn't already do.

A Gamer's Browser?

I took Opera GX for a run on a Nobara Linux 43 setup, using an Early Access version to test it across various scenarios ranging from general browsing to playing YouTube content to running internet speed tests.

On first launch, the browser asked me if I wanted to send telemetry; I declined and moved on to the initial setup. It asked me to pick a theme, and I went with the default GX Classic since it felt more Opera than the rest.

It then asked whether I wanted background music that reacts to my browsing, sound effects for in-browser actions, and keyboard sounds as I typed. I left all of these at their defaults since, honestly, they went right over my head (don't call me a boomer pls).

I could also enable a bunch of sidebar integrations for Telegram, X, Instagram, and so on, but I left those turned off. Opera GX even asked me to import data from another web browser, but it failed to detect Vivaldi, which was already installed.

The next step involved me manually turning on the ad blocker (and later the block trackers option), but the other two features, GX Control and GX Cleaner, were toggled on by default.

I also had to disable Opera AI from the hamburger menu, as I didn’t need it. What did catch my eye, though, were the many pre-installed sponsored speed dials I had to clear out.

And, to little surprise, the default search engine was set to Google, but that is changeable from the Settings menu.

Interestingly, while the ad blocker does work, it fails to show data on how many ads and trackers were blocked when I clicked on the widget for it in the top bar (the shield-looking icon).

I headed into the Privacy and Security section of the settings and found quite a few things enabled by default: automatic sending of crash reports, fetching images for suggested news sources, displaying promotional notifications, and receiving promotional speed dials, bookmarks, and campaigns.

Not great for anyone who takes their privacy seriously and just doesn't want to be bombarded with spammy notifications and speed dial suggestions.

I then fired up Arc Raiders and tried my hand at GX Control. It is Opera GX's built-in resource management panel that lets you cap how much RAM, CPU, and network bandwidth the browser can use with individual toggles and sliders for each limiter. It worked as advertised and even threw up warnings when I set the limits too low.

Similarly, GX Cleaner is the browser's built-in browser cleanup tool that clears out cache, cookies, history, tabs, downloads, and more. It has handy presets, MIN, MED, and MAX, that control how deep the sweep goes, ranging from a light clear of recent temporary files all the way to a full wipe of just about everything. It worked as expected during my use.

A few things that I skipped testing were the stuff most browsers have like bookmarks and extensions, the latter of which Opera GX supports from the Opera catalog.

Then there are the other GX-specific bits, account sync for carrying your data across devices, and the sidebar webapps for Twitch and ChatGPT, which let you keep a stream or an AI assistant open without leaving your current tab.

GX Mods is also there, giving you access to over 10,000 community-made themes, sounds, shaders, and UI tweaks, though you will need an Opera account to get into it.

Wasted?

Depends. For someone like me who tends to close any unnecessary apps running in the background before launching a game, I don't see much use for a gaming browser. For casual browsing tasks, the occasional Alt+Tab to a regular browser does just fine, and the Steam overlay's built-in browser is handy too (albeit very barebones).

That said, if you are the kind of person who RGBs everything in sight and already has a riced-out Linux setup, Opera GX could be a decent addition to the collection.

Just go through the default settings before you do anything else. A lot of what's enabled out of the box won't sit well with most Linux users.

You can grab the DEB and RPM binaries for Opera GX from the official website.