Everyone is wrong about ArtCraft

Summary

The video examines ArtCraft, a project using AI-assisted development to build open-source alternatives to professional creative and productivity software. The presenter argues that many online descriptions exaggerate what the project has achieved.

What the presenter found

• These are not decompiled Adobe apps or exact copies. After reviewing ArtCraft’s documentation, repository, and commit history, the presenter found no evidence that its developers decompiled Adobe binaries or translated proprietary code into Rust. Instead, the project appears to infer how target apps behave through their user interfaces, then implement new code to reproduce that behavior. This is a form of reverse engineering, but not binary-level reconstruction.
• The clones are still works in progress. They are not drop-in replacements for the original programs. For example, PSD files still do not work in the Photoshop clone. ArtCraft estimates that its projects have roughly 40–60% of feature parity, with another 250–1,000 Claude hours needed for each to reach parity. Those estimates have changed little despite substantial development.
• The development process is notable. The team targets specific applications and versions, identifies missing features, implements them, and tests and refines behavior using command-line tools. Separating the UI from the underlying engine helps speed up iteration. The presenter describes the code as largely “vibe-coded,” with many commits and limited depth in early implementations.
• Feature parity involves more than matching visible controls. ArtCraft tracks both feature breadth—the number of features present—and depth, including edge cases and interactions between features. A clone may look convincing during casual testing while still lacking the reliability, performance, format support, hardware testing, and interconnected capabilities professionals depend on. Collaborative apps such as Figma also introduce networking challenges.
• The project demonstrates both AI’s progress and its limits. The presenter considers the speed and polish impressive, especially for a small team, but argues that reproducing established software remains difficult even when the target behavior is known. The project is neither proof that software engineering is “solved” nor evidence that AI-generated applications are useless; its longer-term success remains unproven.

Reviews, guides, or tutorials

No step-by-step guide is provided. The video is a technical analysis based on the presenter’s review of ArtCraft’s code, documentation, and development history.

Main speakers and sources

The video is narrated by Tom Delalande. It also cites ArtCraft’s documentation, repository, commits, and project estimates, alongside online headlines and claims that the presenter disputes.