Rob Shocks
Rob Shocks is a content creator active on video platforms, focusing on artificial intelligence and programming technology. His work centers on evaluating AI coding tools, sharing usage experiences, and analyzing technology trends. Targeting developers and technology enthusiasts, he presents the practical applications and capability boundaries of AI tools from a hands-on testing perspective.
Timeline (last 90 days)
Rob recommends that the level of ceremony should match the level of risk: change a button color directly, use grill me plus TDD for small features, use a spec and phased plan for large or vague features, and reserve full review only for production-safety or irreversible operations.
In Rob's Vault application test, after the agent opened the binary with Ghidra it encountered more than 242 anonymous functions with stripped names, then within a few calls located the key validation function, reconstructed it into C-like code, and wrote a verification-code generator.
In reverse engineering, the fact that an agent generates entirely new code after understanding the logic—rather than copying the original code—is regarded as the essence of reverse engineering.
Rob believes that code has never been a true moat, and that real barriers lie in data, users, cloud-side security, delivery speed, taste, and things running on one's own servers that cannot be inspected.
If you publish a desktop application, browser extension, game, mobile app, or client-side JavaScript, developers should assume that any person's AI agent can figure out how it works.
Reverse engineering performed for interoperability purposes is generally legally protected, but many applications' terms of service prohibit reverse engineering, and bypassing DRM or encryption is a separate legal issue; the safe zone consists of your own applications, open-source binaries, security challenges, and targets you are authorized to test.
All Facts (6)
Rob recommends that the level of ceremony should match the level of risk: change a button color directly, use grill me plus TDD for small features, use a spec and phased plan for large or vague features, and reserve full review only for production-safety or irreversible operations.
50%UnverifiedIn reverse engineering, the fact that an agent generates entirely new code after understanding the logic—rather than copying the original code—is regarded as the essence of reverse engineering.
50%UnverifiedRob believes that code has never been a true moat, and that real barriers lie in data, users, cloud-side security, delivery speed, taste, and things running on one's own servers that cannot be inspected.
50%UnverifiedIf you publish a desktop application, browser extension, game, mobile app, or client-side JavaScript, developers should assume that any person's AI agent can figure out how it works.
50%UnverifiedReverse engineering performed for interoperability purposes is generally legally protected, but many applications' terms of service prohibit reverse engineering, and bypassing DRM or encryption is a separate legal issue; the safe zone consists of your own applications, open-source binaries, security challenges, and targets you are authorized to test.
50%UnverifiedIn Rob's Vault application test, after the agent opened the binary with Ghidra it encountered more than 242 anonymous functions with stripped names, then within a few calls located the key validation function, reconstructed it into C-like code, and wrote a verification-code generator.
50%