A new platform called REA allows developers to inspect software behavior and reconstruct its logic using natural language prompts. The tool analyzes compiled programs, extracts operational rules, and provides decompiled code and execution context for verification and modification.

The platform functions by connecting to a running application or debugging endpoint, then retrieving source-level artifacts such as instruction traces, configuration constants, and control flow patterns. Once gathered, these elements are presented to an AI agent that interprets them in plain language. Users can then request explanations, rebuild logic, or simulate modified behavior in isolated test environments.

In one documented case, REA was used to analyze Windows Calculator’s percentage function. The investigation revealed that the % operator applies different mathematical rules depending on the preceding operation. After addition, it calculates 10% of the first operand (200 × 0.1 = 20), then adds that value to the first number (200 + 20 = 220). After multiplication, it treats the percentage as a multiplier of the first operand (200 × 0.1 = 20), which is then applied in subsequent calculations. This behavior was confirmed by matching assembly-level instructions to known operator codes in the application’s library.

Another example involved a browser-based dinosaur game. By inspecting the game’s runtime script, REA identified that speed increases by 0.001 per update cycle until reaching a maximum of 13, starting from an initial value of 6. This rule was then replicated in a minimal standalone game with an adjustable speed slider, allowing users to experiment with the same acceleration logic.

REA supports multiple environments, including native Windows executables, Electron applications, and browser-based JavaScript bundles. For native binaries, it traces function calls and memory references; for web-based apps, it captures module loading and runtime state. The platform does not modify original software but enables reproducible analysis and synthetic recreation under controlled conditions.

The tool is designed for technical users seeking to understand, audit, or extend software without access to source code. Its workflow emphasizes factual reconstruction over speculation, relying on observable behavior and documented code patterns. Developers can use it to reverse engineer functionality, validate assumptions about algorithmic rules, or build educational demos based on real implementations.

While REA does not name specific commercial adopters, its methodology aligns with practices used in software preservation, security research, and interactive system design. The platform’s documentation includes setup instructions for connecting to local debugging endpoints and parsing executable artifacts.

The ability to translate low-level behavior into high-level explanations without requiring deep expertise in assembly or binary formats represents a shift in how developers interact with compiled systems. Rather than manually tracing instructions, users can query the system directly and receive structured, evidence-based responses grounded in actual program execution.