Software Is Dead, Long Live Software

For decades, software companies have operated on a relatively simple economic principle: developing complex applications takes time, money, and expertise. Once built, those applications can be distributed to millions of customers at almost no additional production cost. The expensive part is creating the software, not copying it.

That helped establish an industry worth hundreds of billions of dollars. Companies such as Adobe, Microsoft, and Autodesk built businesses around applications that would be difficult and expensive for competitors to reproduce. AI is now lowering that barrier. Work that once required large engineering teams can increasingly be attempted by much smaller groups, sometimes even individual developers.

In October 2026, developer Brandon Thomas drew attention with ArtCraft, a collection of seven free, open-source applications intended to recreate much of Adobe's Creative Cloud. Built with help from Anthropic's Claude and written in Rust, the projects target Photoshop, Illustrator, Premiere Pro, Lightroom, After Effects, InDesign, and Acrobat.

The applications are rough around the edges. They're slow, missing important features, and nowhere near ready for professional use. But the fact that one developer could get seven recognisable alternatives this far with AI assistance is worth examining.

Rebuilding an empire

Adobe's Creative Cloud represents decades of engineering. Its applications contain specialised tools, compatibility features, performance optimisations, and workflows refined through years of professional use. Reproducing that ecosystem would traditionally demand a substantial development effort.

ArtCraft's developer says the applications were built without access to Adobe's source code. Instead, AI helped recreate how the programs look and behave. Their interfaces resemble Adobe's, including familiar panels and tools, although resemblance doesn't guarantee compatibility.

Tests published by Creative Bloq found problems with image processing, inaccurate masking, missing features, and extremely slow 4K video editing. FilmCraft couldn't open Premiere project files, while PhotoCraft struggled with tasks Photoshop handles routinely. These shortcomings matter, especially to professionals who depend on their software every day.

Still, getting a recognisable starting point into people's hands is becoming easier. If AI continues to reduce the cost of that first implementation, established vendors may find it harder to justify prices solely on the grounds that competing software is expensive to build.

The declining cost of software

Modern AI coding tools can turn descriptions into executable code, generate interfaces, write tests, and help developers debug problems. They don't remove the need for engineering judgement, but they can change how much manual work goes into a new application.

Suppose building an alternative to a €100-a-year commercial tool would traditionally cost €10,000. If AI cuts the engineering work by roughly two-thirds, that initial investment becomes more manageable. Spread the development and maintenance costs across a thousand users, and the economics can start to favour a shared open-source project.

Those figures are hypothetical, and maintenance, security, and support still cost money. The point is that a company may continue to own its source code while losing some of the pricing power that came from how difficult it was to recreate the same functionality.

There's also a distinction between copying a program and building another program that does similar things. Software can perform the same calculations, read compatible files, or provide familiar workflows using different code. In a 2012 case involving SAS Institute and World Programming, the Court of Justice of the European Union held that software functionality, programming languages, and certain data-file formats were not themselves protected as expressions of a computer program under the applicable copyright rules.

That judgment doesn't give developers unrestricted permission to reverse engineer anything they like. Access restrictions, protected code, and contractual obligations can still complicate independent projects. But a competitor doesn't necessarily need the original source code to build a useful alternative.

For users, this could mean more choice across operating systems and less dependence on a single vendor. A Windows-only application could inspire a Linux-native counterpart. An abandoned commercial program could gain an independently maintained successor. A proprietary format might become usable in more than one application.

Getting the details right

Generating code quickly isn't the same as producing software people can trust. AI-assisted programming research has produced mixed results: some controlled studies report productivity gains, while a 2025 study of experienced open-source developers found that participants took about 19% longer with AI tools on familiar codebases.

Writing the first version is only part of the work. Developers still need to test it, fix bugs, maintain compatibility, handle unusual inputs, and address security problems. ArtCraft's small installation sizes and native Rust implementations haven't automatically made its applications fast or dependable.

Open source does offer a route for improving unfinished projects. Other developers can inspect the code, replace weak implementations, add missing features, and investigate performance problems. Whether enough contributors will join ArtCraft to make that happen is another question. As generating an initial version becomes cheaper, the harder and more expensive work may increasingly lie in verification and long-term maintenance.

What Adobe may have to sell next

Adobe has advantages that are difficult to reproduce. Its applications work together, professionals know their interfaces, and businesses rely on them for projects where reliability matters. Cloud services, established file formats, support, and a large user base give customers reasons to stay even when free alternatives exist.

As competing applications become cheaper to develop, however, software companies may have to reconsider what customers are paying for. Reliable infrastructure, security, support, collaboration, and compatibility could matter more than access to a particular set of features. For Adobe, that might mean more flexible subscriptions, better interoperability, and fewer obstacles for customers who want to work across different tools.

Smaller vendors could feel the pressure sooner. File converters, administrative utilities, and specialised productivity tools are generally easier to recreate than complex enterprise platforms. Some charge recurring fees for fairly straightforward functionality. They are natural candidates for AI-assisted alternatives, especially when users can share the cost of maintaining them.

The way people use software could change as well. Until now, users have largely adapted their work to the applications available. If development becomes affordable enough, a small business might commission a tool tailored to its own workflow, while an open-source community could keep an abandoned application alive. People would have more opportunities to shape the software around their needs rather than the other way around.

ArtCraft hasn't displaced Adobe, and its current limitations show how much work separates a prototype from a dependable product. But it points to a market where building a recognisable competitor may become progressively cheaper. Established companies will still benefit from their experience and ecosystems, though those advantages may not always justify the prices customers have grown used to paying.

The software industry isn't going anywhere, but the economics that shaped it for decades are changing. As reproducing functionality gets cheaper, companies will need to give customers better reasons to keep paying. Owning the code might no longer be enough.