The Adobe Clone That Went Viral Before It Was Ready #
This week, a developer named Brandon launched ArtCraft, an open-source suite of Adobe-inspired applications, sparking immediate viral excitement. His stated motivation: deep frustration with Adobe’s “dark pattern” subscription cancellation fees, a grievance many creatives share.
Brandon claims a clean-room approach, meticulously rebuilding the seven applications—including PhotoCraft, VectorCraft, and FilmCraft—in Rust. He asserts that no Adobe code or assets were incorporated; instead, he reverse-engineered public specifications and observed application behavior, even writing PSD support from Adobe’s public file format spec. However, ArtCraft’s familiar interfaces, nearly identical to Adobe’s, inevitably raise concerns about intellectual property copying.
Despite the initial euphoria, early testing on Hacker News quickly tempered expectations. Users reported misaligned text selection, non-working shortcuts, unresponsive font size sliders, and non-collapsible folders, with one tester calling it "actual joke software." ArtCraft’s own roadmap, under an “honesty parity assessment,” concedes that Photoshop parity currently sits well below 50%, with 99% parity projected to take months, not years.
A Repo That Gives Coding Agents X-Ray Vision #
Brandon’s project is a “clean-room rewrite,” a painstaking, manual process. But a different viral GitHub repo, Reverse Engineer Anything (REA), suggests a more automated path to app cloning. This isn't a magic button; it's an MCP (Model Context Protocol) server and CLI connecting AI agents to industrial-grade binary analysis tools like Ghidra, Hopper, and IDA Pro.
REA grants agents X-ray vision across diverse software: native binaries (Mach-O, ELF, PE), Electron apps, .NET assemblies, Android APKs, and even web applications. An agent can trace a feature, citing specific code evidence, then implement a similar version for your project. This transforms inspection and decompilation into practical, actionable coding workflows.
Consider the detailed showcases: agents replicate Notion’s clipboard feature or rebuild a retro game’s sound from C. While setup still requires installing MCP servers and tools, and verification remains critical, REA dramatically shifts the effort. It moves beyond speculative “clean-room” rewrites to targeted, evidence-based feature extraction and implementation, hinting at a future where proprietary code becomes a mere blueprint for AI-driven replication.
When a Clone Takes Two Hours—and $140 #
Stork.AI’s own models have proven the point. One experiment tasked Opus 5.5 with reverse engineering the 1995 game Space Cadet Pinball from its Windows executable, then recreating it natively in Rust for macOS. The agent disassembled the game, wrote its own tools, and translated it function-by-function, splitting the work across 15 sub-agents for physics, flippers, and 32 missions.
Two hours and 25 minutes later, it had generated 31,000 lines of Rust, 174 tests, and zero dependencies, all for roughly $140 in API costs. While impressive, this isn't proof that any complex application can be cheaply recreated. A bounded game, with clear pass/fail conditions for behavior, is far easier to test than a sprawling creative suite like those found in the ArtCraft GitHub Organization.
Agents excel when behavior can be repeatedly checked against a known good state. This is precisely why test-driven porting efforts, such as VNext (porting Next.js to Vite) or the TypeScript-to-Rust compiler effort, have seen success despite substantial compute costs. Agents simply keep working until the tests pass, demonstrating that verifiable code generation is their true strength.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
Software’s Moat Is Moving—and the Law Still Applies #
Software’s traditional moat is shifting: client-side behavior becomes increasingly inspectable, but a product’s true value extends far beyond its UI. Reproducing a complete application means replicating its reliability, cloud services, data integrity, third-party integrations, and complex enterprise requirements—a task far more demanding than a functional recreation.
Legal uncertainty also looms over these cloning efforts. While clean-room functional reimplementation has precedent, direct copying of proprietary code, copyrighted assets, branding, or UI trade dress invites separate, significant legal risks. Patents on specific algorithms or features could also create separate infringement claims.
The broader consequence is clear: proprietary code is a weakening moat. Backend infrastructure, robust cloud services, and established ecosystems now define a product’s defensibility. This shift makes personal software more achievable for individual users, even as public, professional-grade clones struggle to reach parity.
Frequently Asked Questions #
What is ArtCraft?
ArtCraft is a suite of open-source apps designed to reproduce parts of Adobe products, including Photoshop-like and video-editing tools.
What does Reverse Engineer Anything do?
It connects AI coding agents to tools such as Ghidra, Hopper, and IDA to inspect software and help explain or reproduce its behavior.
Can AI fully clone professional software?
AI can reproduce features and workflows, but matching reliability, integrations, polish, and professional parity remains difficult.
Is reverse engineering software legal?
It depends on jurisdiction, method, licenses, and what is copied. Clean-room implementations may be treated differently from copying protected code or assets.