The Foxes Are Running the Hen House

Enshittification is a three stage process: "First, platforms are good to their users. Then they abuse their users to make things better for their business customers. Finally, they abuse those business customers to claw back all the value for themselves." Customers watch the OS get worse. Developers hand over 30% of every sale to distribute their apps. Nothing meaningfully changes.

The obvious response is regulation. The EU's Digital Markets Act forced Apple to allow third party app stores. Apple complied by making the alternative so hostile and fee-laden that it functionally does not exist. Third party stores are region-locked. The onboarding process is deliberately friction-heavy. Apple still charges a toll to use the door they were legally required to unlock, and they spent seven million euros lobbying to avoid unlocking it in the first place. Google is running the same play. This is not a company outrunning government. This is a company inside the regulatory process, funding it, contesting it, and shaping the implementation until compliance produces the appearance of change while the chokepoint stays intact.

Apple, Google, and Microsoft are trillion dollar companies. The legal teams contesting DMA compliance dwarf most regulatory agencies in budget and headcount. You cannot out-lobby them. You cannot out-lawyer them. You cannot out-pizza the hut.

But there is a deeper reason regulation keeps failing. Cory Doctorow says the piece of received wisdom he would most like to abolish from tech policy is this: "Tech moves at the speed of innovation and regulation moves at the speed of government, so regulation will always lag behind tech." The problem is not speed.

Regulation only has teeth when the people it is meant to protect have somewhere else to go. Apple knows developers cannot abandon iOS users. Google knows app developers cannot walk away from Android. Right now Linux on desktop and mobile is not that escape hatch.

The largest problem is that there is no development target. As a company, saying you support Linux is not a statement. It is a question. Does that mean Fedora? Ubuntu? Arch? Older versions? Newer versions? X11 or Wayland? GNOME or KDE? Systemd or not? Et cetera, et cetera, ad infinitum.

But even if you pick one distribution and commit to it, you have not solved the problem. You have only narrowed it. Two machines running the same version of the same distribution can have completely different OS layers underneath your software: different packages installed, different runtime versions, different configurations accumulated over years of different decisions. There is no such thing as a canonical Ubuntu 24.04 installation.

A customer files a ticket. You have no way of knowing whether the bug is yours or theirs. On Windows you debug the software. On Linux you debug everything.

The server world solved this problem twenty years ago. Containers. Reproducible images. Identical environments that behave the same on every machine, every time. Cloud infrastructure is built on one principle: every server is the same server. That is why Linux won the cloud. Not just because it was free. Because it became a platform you could reason about.

Linux powers 96.3% of the top one million web servers. 100% of the world's most powerful supercomputers. 92% of virtual machines across AWS, Azure, and Google Cloud. The entire modern internet, the AI infrastructure, the cloud platforms that Apple and Google run their own businesses on: it all runs Linux.

This did not happen because of ideology. It happened because of one structural fact: no toll booth. Windows Server licensing starts at $1,176 for a 16-core licence plus Client Access Licences for every user who touches it. Linux is free. At the scale of millions of virtual machines and billions of container starts, that is not a marginal difference. It is the reason the market moved.

The retail industry ran the same calculation. NCR Voyix builds the point-of-sale terminals in supermarkets and restaurants globally. Per-device Windows licensing across 10,000 checkout terminals is economically ruinous. Remove the toll booth and the market moves.

This is the model. Remove the toll booth, build the infrastructure, and the market moves. It does not move because of values or ideology. It moves because the economics change. The server took fifteen years from curiosity to 92% of the cloud. The desktop and mobile are next.

The Linux desktop is now doing what the server did. It is taking the same principles that won the cloud: reproducible images, identical environments, no configuration drift. Those solutions exist. They are shipping now.

Image-based immutable desktops deliver the OS as a single read-only image rather than a collection of individually managed packages. GNOME OS, Bluefin and its experimental Dakotaraptor variant, Endless OS, and the forthcoming KDE Linux are the leading examples. Every device running the same OS version boots from an identical, cryptographically verified base image. That image cannot be modified at runtime. Applications live above it in sandboxed containers. When an update ships the entire image is replaced atomically on the next reboot, with instant rollback if something goes wrong.

The support ticket that was impossible to diagnose on traditional Linux becomes straightforward. The base is identical on every machine. If the bug reproduces, it is your code. If it does not, it is their machine.

They also share a common application runtime. Software packaged against the Flatpak runtime, specifically the Freedesktop SDK that these OSes build on, runs identically across all of them. Ship once against the runtime, run anywhere the runtime exists. When a fix ships at the OS layer it is delivered as a new image and applied uniformly across every installation. There is no configuration drift, no user stuck on a library version that predates the fix.

This gives developers something they have never had on Linux before: a certification target. You can now say your software is certified for Freedesktop SDK 25.08, and that statement carries the same weight as when a mobile developer says "requires Android 14 or above" or "requires iOS 15 or above." Any OS built on that SDK version is a known, stable, predictable environment your software will run on. Not "tested on Ubuntu 26.04 and Fedora 44, results may vary elsewhere." Certified.

This matters beyond developers. Right now, a government that runs its ministries on US software cannot take a position that conflicts with US interests without risking its own infrastructure. That is not a hypothetical. When the International Criminal Court issued an arrest warrant for a US ally, the Trump administration had Microsoft shut down the court's IT systems. The chief justice lost access to his email, his working files, his calendar. One phone call to Seattle.

A government running on open infrastructure cannot be switched off that way. That is what digital sovereignty actually means. Not a branding exercise. An answer to a specific question: who has the ability to shut you down?

The leverage runs in both directions. A government that depends entirely on three US companies for its computing infrastructure cannot credibly threaten to regulate those companies. The companies know it. The governments know it. A viable open platform changes that calculation. You cannot credibly fine someone you cannot afford to lose.

This is an unknowing coalition built on free market dynamics. App developers paying 30% today with no leverage against 40% tomorrow. EU governments running Linux across their server infrastructure while the software their citizens use daily depends on US platforms. National security establishments who have watched a US president use technology infrastructure as a foreign policy instrument and drawn the obvious conclusion. Hardware vendors who want to ship a device and own the customer relationship without paying certification fees to a platform owner who is also their competitor. Digital rights advocates who have been making this argument for twenty-five years and are finally being joined by people who want to get rich rather than prove a point. Users on every platform who benefit from competition existing even if they never switch. Nobody signed up for this coalition. The foxes assembled it by building toll booths everywhere.

Two things are preventing the platform from tipping. Neither is technical.

1. Governments and Businesses decisions are critical

Every technology decision is also a platform allegiance decision. Most of the people making them do not know that.

When a government procurement team chooses a verification system that requires a Google-certified device, they are not only making a security decision. They are deciding which devices their citizens are allowed to use to access public services. When a development team integrates App Attest because it is well-documented and widely supported, they are not only authenticating users. They are embedding Apple's market position into their product. These choices do not feel like platform decisions. They feel like implementation details. That is why they keep getting made.

The path of least resistance is to use the Google verification SDK because it is well-documented and widely used. The path of least resistance is to mandate Play Services for a civic app because the security team is familiar with it. The path of least resistance is to add an anti-tamper check that happens to flag Wine as a modified environment. Each decision is defensible in isolation. A procurement manager who chose the well-documented option over the obscure one made a reasonable call. A security team that required device certification for a public health app was doing their job.

The problem is not the individual decisions. It is what they add up to. Every platform-captive choice makes the alternative harder to use. Harder to use means fewer users. Fewer users means fewer developers justify building for it. Fewer developers means the alternative shrinks. And a smaller alternative means the platform owners have more leverage, not less. The institutions most loudly concerned about digital sovereignty are the ones most consistently making it harder to achieve.

None of this is malicious. Most of it is thoughtless. The consideration that is missing is simple: does this decision foreclose an exit option that someone downstream might need?

2. The Chicken and the Egg

Users will not move to Linux if the apps are not there. Developers will not ship on Linux if the users are not there.

No individual developer can break this loop alone. If you ship your app on Linux and nobody else does, your users do not follow you there and you have done expensive work for nothing. The move only pays off once enough developers have moved that users have a reason to switch. But developers are waiting for the users, and users are waiting for the developers, and so nothing moves.

This is the chicken and egg problem. The solution is not to convince every developer simultaneously. It is to get enough of them to move that the platform tips.

The flywheel does not need to spin fast to change the power dynamic. It just needs to start spinning. Every app that ships on Linux makes the next developer's decision easier. Every user who moves to Linux because their apps are there is a user the platform owners can no longer take for granted. The math changes slowly and then quickly, the same way it changed on the server.

Photoshop on Linux would be the moment. Not because Linux users need Photoshop specifically. Because Photoshop on Linux tells every professional software developer that this platform is serious and that their customers can get there. It changes what every developer can demand from Apple and Google in their next negotiation whether those developers use Linux or not.

It also unlocks an entire category of professional user that is currently locked out. Design studios run on Adobe Creative Cloud. That dependency is the single biggest reason creative professionals and the businesses employing them cannot seriously consider Linux as a primary platform. Not the terminal. Not the package manager. Not the desktop environment. Photoshop, Illustrator, and Premiere. Remove that blocker and design studios can evaluate Linux on its actual merits: no per-seat OS licensing, no platform owner extracting a cut on every software purchase, full control over hardware and update cadence. Money that currently flows to the platform owner stays with the business.

That is what a single high-profile port does to the flywheel. It does not just add Linux users. It makes the exit credible for entire professional categories.

This is the Shopify moment for creative professionals. Shopify did not kill Amazon. It gave retailers an exit option, and that option alone shifted the negotiation. Amazon had to compete for sellers it previously took for granted. A viable Linux ecosystem for design studios does the same thing to Apple and Microsoft. They have held pricing power and dictated platform terms because leaving meant abandoning your entire workflow. The moment that is no longer true, the calculus changes. Not because studios leave en masse. The threat becomes real. Real threats change behaviour.

Each category that opens up makes the platform more credible to the next developer considering a port. The loop that has kept Linux at five percent market share runs in both directions.

The minimum ask is not that you ship on Linux. It is that you stop making it harder. Do not patch your software to break compatibility tools. Do not adopt platform-captive verification when a neutral alternative exists. Do not treat Linux users as second-class customers. The Linux community will do the heavy lifting of making your software run if you give them the room to do it.

The larger contribution is to show up. Ship something, even a beta. Your presence is the data point the next developer uses when they decide whether this platform is worth their time. Coordination problems are solved by enough people deciding to coordinate. The foxes built the henhouse. They staffed it with your apps. They are raising the rent. Build one they do not own.