A developer says he has used an AI model to build open-source replacements for seven of Adobe's core creative applications — and the repositories are now climbing GitHub's Trending page.

The apps were announced by the team behind Artcraft, an AI image-and-video studio, and reimplement Adobe's best-known tools in Rust under MIT/Apache licences, with WebAssembly builds that run in a browser: PhotoCraft (Photoshop), VectorCraft (Illustrator), FilmCraft (Premiere), LightCraft (Lightroom), EffectCraft (After Effects), DesignCraft (InDesign) and PdfCraft (Acrobat Pro).

In posts on Reddit and Hacker News, developer Brandon Thomas said he used Anthropic's Claude Opus to produce "clean-room replacements" — reimplementations that reproduce a program's behaviour without copying its source code. Thomas, who says he has about a decade of experience on large-scale Rust projects and has coded exclusively with AI tools since February, told Ars Technica he was motivated by being "bit by the dark pattern 'cancellation fee' one too many times."

The software is genuinely free: the code is open, and Thomas says development is funded by selling token-based access to the pre-existing Artcraft AI studio, which appears inside the apps as one image-generation option alongside third-party models.

The claims are as large as the release. Thomas said the current "super early alpha" state would reach full feature parity "within a month" — then softened the estimate on Hacker News to "99% parity will take a while, but I'm sure it'll be measured in months and not years." In one comment he declared "Software is over," adding that he had "one-shotted Photoshop. The whole thing." He has predicted a near future in which anyone can recreate 1:1 functional equivalents of paid software to "take back the internet and own tech forever."

Reality is less tidy. Ars Technica described the apps as "ambitious, free, and nowhere near finished," and commenters on Hacker News and Reddit have catalogued missing tools and rough edges — a normal state for a project weeks old, but a long way from parity with a suite Adobe has polished for decades. The apps also lean heavily on visual similarity to Adobe's interfaces; the same Ars piece notes that could expose them to "trade dress" claims, even though reverse-engineering functionality has generally been found legally acceptable.

Why it matters: whatever the maturity of the code, this is the clearest demonstration yet that one developer with an AI agent can attempt to reproduce an entire commercial software suite — and ship it. It also raises a question clean-room practice has never had to answer: the method traditionally requires strict separation between the people who write a specification and the people who implement it, so implementers are never tainted by the original code. When a single large language model does both ends of that job, the separation — and the legal reasoning built on it — becomes far murkier.

Sources