# Your Vibe-Coded App Can Look Right And Fall Apart After Launch. REA Shows You How Real Teams Build.

## Executive summary

The video introduces REA (Reverse Engineer Anything), a tool that integrates with AI agents (like Claude Code or OpenAI Codex) to allow users to analyze existing software applications at a deep, technical level. For build engineers, REA acts as an 'x-ray machine,' bridging the critical gap between high-level business requirements and the complex, functional code implementation. It enables both non-developers and seasoned engineers to understand the underlying logic, data structures, and engineering decisions of existing systems, facilitating the creation of new, robust software without needing to replicate the entire original architecture.

## Key takeaways

- REA's Purpose: Understanding vs. Cloning: REA does not promote software death; rather, it provides a way to understand how real teams translate vague 'wants' into working code. It helps users learn from existing implementations rather than simply cloning them.
- Non-Developer Advantage: For non-developers, REA serves two key functions: 1) Understanding the necessary components of the desired software, and 2) Understanding *why* an existing implementation made specific engineering choices, allowing the user to select the most appropriate patterns for their own build.
- The Power of Contextual Analysis: Analyzing a single feature, like the percentage button in a calculator, reveals that functionality is highly dependent on context (e.g., addition vs. multiplication), a detail that high-level requirements often miss.

## Technical details

- Authenticator App Logic: REA investigated Aegis, an Android authenticator app, revealing that the six-digit code is generated by combining an account key with a current time, grouped into 30-second intervals. The code's consistency is dependent on both the key and the time interval.
- Notion's Clipboard Mechanism: The investigation into Notion's copy/paste process showed that the application sends text and HTML to the clipboard but also maintains structured block data locally. Crucially, it embeds an identifier in the HTML, allowing the paste function to reconnect the visible content with the saved structured data.
- QuickBooks to Independent App (Ledger Reborn): A developer used REA to investigate QuickBooks Desktop for Mac, focusing on inventory arithmetic and reconciliation balances. This investigation provided the necessary evidence and pseudo-code to build an independent accounting application (Ledger Reborn) using Python and SQLite, demonstrating that the investigation informs new engineering choices.
- Browser Game Logic (Dinosaur Game): REA inspected a Chromium-derived dinosaur game and retrieved the JavaScript logic for acceleration. The code dictates that the speed increases by a thousandth on each update, providing a precise rule for the game's feel that can be used to implement a new, controlled speed mechanism.

## Practical implications

- The process of building software shifts from merely defining requirements to actively interrogating existing systems to discover necessary technical requirements (e.g., how relationships between data objects must be preserved during copy-paste).
- Developers can use REA to narrow down complex feature implementations, asking specific questions like, 'How exactly did they get that behavior?' rather than needing to reconstruct the entire system.
- The tool empowers non-technical users to participate in an 'engineering conversation' by asking agents to interpret code and explain dependencies, making them valuable contributors to the design process.

## Topics

Reverse Engineering, Software Architecture, AI Agents, Requirements Engineering, Data Modeling, Build Process, REA (Reverse Engineer Anything), Claude Code / OpenAI Codex, Aegis, Notion, QuickBooks Desktop

Source: https://www.youtube.com/watch?v=PseF3LyXQ04