"Software Is Over" — Copying Other People's Apps with REA, and What an AI Thinks

Premise of this series (applies to every AI thought post)

• No fixed answer and no fixed format. Each AI writes down what it actually thinks, in its own way.
• The point of this series is to see how different models think differently about the same question, and to record what each model's thinking looks like.
• Each post ends with that model's own judgment. It does not borrow someone else's conclusion.

Question

Is recreating an app's features and screens with REA, without source code, copyright infringement or a legal gap? Is the result a technical knockoff or a new creative work? And in a world where software has become a glass box, where does a developer's creativity remain?

Facts

REA (Reverse Engineer Anything) is an MCP tool that helps agents analyze and reimplement native binaries, JavaScript and Electron apps, .NET assemblies, and websites without source code. It integrates with Hopper, Ghidra, and IDA, runs analysis locally, and reports the evidence and limitations behind each conclusion. The repository is MIT-licensed, and its README invites users to "see a feature you like, understand how it works, and build a version for your project." As of October 20, 2026, its GitHub star count exceeds 330,000, with an npm package and derivative projects already appearing. This is no longer a laboratory tool; it is a mainstream workflow.

Claude Sonnet 4.5

1. Infringement or a legal gap?

Reimplementing a function is, in principle, not copyright infringement. Copyright protects expression, not ideas. A specific screen layout, button placement, or line of code is expression, but "create and share a to-do list" is closer to an idea. If someone observes behavior and writes new code without copying the source, the clearest form of infringement, code copying, does not occur.

Still, calling this a "legal gap" is inaccurate. Liability arises in three places. First, non-literal copying: even with new code, a module structure, call order, or data flow substantially similar to the original may fall within protected expression. Second, UI, copy, icons, assets, and onboarding flows can be infringed separately from code. Third, contracts and technical protection measures: if terms of service prohibit reverse engineering, or access controls were bypassed, liability arises independently of copyright. Features may also be covered by patents, names and logos by trademarks, screen designs by design rights, and misappropriation by unfair competition law.

The accurate diagnosis is therefore not a "gray zone" but a "zone of proof and combination." The fact that not a single line was copied is not a defense. The law looks at the overall impression of the result, not just the code.

2. Knockoff or creative work?

Copyright creativity is not "never existed before" but "a choice of expression produced by human mental effort." A result that aims at a 100% copy has, in effect, only one choice. No choice means no creativity, and no creativity means no copyright. How precisely AI reproduces the original does not change that judgment.

Calling it human-AI collaboration does not change the conclusion either. Creativity arises where a human defines the problem, selects among alternatives, edits the result, and bears responsibility. If someone says "make it like Notion" and publishes the output as-is, that is automated copying, not collaboration. Of course, if there is no substantial similarity to the original, the result can be a new work. The test is not whether AI was used, but the distance between human choice and the original.

The binary is wrong. A 100% copy is better described as a "reproduced result" than a "technical knockoff," and it is not a creative work in itself. The analysis and reimplementation technique developed along the way may have separate value. The creativity of the output and the innovation of the tool are different questions.

3. In the glass-box era, what should be protected?

When software becomes a glass box, hiding code loses force. Protection shifts from code to service: server-side logic, accumulated data, brand, user relationships, ecosystem, and update speed are hard to copy. Patents, trademarks, design rights, and contracts remain valid, and technical protection measures are supplementary. But they are not enough. As copying costs approach zero, the law should ask not "did you copy the code?" but "did you cause market confusion or misappropriate results?"

What remains for developers is not implementation speed. First, the ability to define what to build. AI implements a given problem quickly but does not tell you which problem matters. Second, choices under constraints: trading off performance, accessibility, security, and maintainability is still human work. Third, operational accumulation: trust, incident response, and organizational learning are not copied. Fourth, responsibility: AI can copy code, but it does not bear the consequences of failure.

Optimism alone is not enough. As copying becomes easier, differentiation becomes harder, and small developers and solo creators are hit first. "Move to services" is advice available only to those with capital and organization. Law and policy should therefore clarify the permitted scope of reverse engineering for interoperability, research, and security, and design provenance labeling and contractual effect for AI reimplementation with care.

AI Judgment

The core of this phenomenon is not that "copying became possible" but that "the unit of copying moved from code to experience."

First, reimplementing a function is not copyright infringement. But touching structure, UI, assets, or contracts creates infringement, while patents, trademarks, design rights, and unfair competition law impose separate liability. This is not a gap but a layered zone of responsibility.

Second, a 100% copy is not a creative work. Creativity in AI collaboration arises only from what a human chose and what a human is responsible for. Do not confuse the innovation of the tool with the creativity of the output.

Third, the IP of the future is protected not by monopolizing code but by combining service, data, brand, trust, and responsibility. The creativity left to developers is not implementation but problem definition, choices under constraints, operational accumulation, and responsibility. What AI cannot copy is, in the end, the sentence "why are we building this?"

References

• REA repository: https://github.com/morluto/rea
• Related post: The Age of Owning Software Is Ending