🔍 Read the full analysis: How To Get Oriented In Someone Else’s Software With AI on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A report published by ThorstenMeyerAI.com on October 19 describes how the open-source Reverse Engineer Anything (REA) tool lets AI coding agents inspect existing programs and investigate specific behaviors. The source presents a QuickBooks Desktop recreation as a case study, but its reported accounting matches and other project metrics have not been independently verified here.
A report published by ThorstenMeyerAI.com on October 19 describes how developers can use the open-source Reverse Engineer Anything (REA) project with AI coding agents to investigate how existing software behaves. The guide focuses on asking bounded questions, checking that REA tools are actually connected and distinguishing observed behavior from an agent’s inferences—steps that can help developers write more dependable software specifications.
The report presents REA as a set of tools for agents including Claude Code, Codex, Cursor and Gemini CLI. It describes a local MCP server, a command-line utility and an agent skill that guides an investigation. MCP is a way for agents to call external tools. REA’s analysis options, as described in the report, cover several types of software and data, including native applications, Electron and JavaScript programs, .NET and Android apps, firmware, saved network traffic, WebAssembly and SQLite databases.
The guide advises starting with one specific behavior and the decision it affects, rather than asking an agent to explain an entire program. Its example is tracing how a partial invoice payment changes the remaining balance and which account receives the posting. The report recommends asking for stored data, calculations and conditions that alter the result, while labeling statements as observed, inferred or unknown. A useful check, it says, is whether the agent can explain a concrete example using inputs and outputs.
Setup instructions in the source include running npx rea-agents@latest setup, approving the proposed changes and checking prerequisites with the project’s doctor command. The report cautions that this diagnostic checks files and prerequisites, not whether an agent has a live connection. It recommends verifying tool access inside the agent before relying on answers. The source says REA requires Node.js 22.19 or newer; some native-program investigations also require a disassembler such as Ghidra, Hopper or IDA.
Infographic 01 · Reverse Engineer Anything
Five places to point REA first, and how to tell when the answer is wrong.
REA gives coding agents such as Claude Code and Codex tools to inspect programs, apps and browser sessions. Each investigation needs a finishable question and a check.
npx rea-agents@latest setupThe five investigations
What to ask, and the check that catches a wrong answer.
- 1
Ask a question you can finish
One behaviour, one decision it informs. Example: how a partial payment changes an invoice balance.
It’s wrong ifThe agent can't walk through one concrete example with real inputs and outputs. - 2
Get REA connected and confirmed
Run the guided setup, then doctor, then call binary_session inside the agent before any real target.
It’s wrong ifAnswers arrive without a single REA tool call: you're reading the model's guesses. - 3
Turn a business rule into a spec
Inputs, stored fields, formula, edge cases, plus test cases checked against the original's own output.
It’s wrong ifYour version and the original disagree on a test case, even by one cent. - 4
See what has to survive a copy
Find what an app writes to the clipboard and what it keeps locally, then design your own copy and paste.
It’s wrong ifThe explanation doesn't predict what lands in a plain text editor, an email and your app. - 5
Your own app, your equipment, an old program
Start with software you own or rely on. Ask what it sends, stores or calculates when you do one thing.
It’s wrong ifIt only works for the example you watched. Test with a new file or a read-only command first.
Sources: REA on GitHub (stars via GitHub API, 19 Oct 2026); REA installation guide; REA security policy.
Powered by Thorsten Meyer AI
From Screens to Software Rules
The approach addresses a common gap in AI-assisted software development: reproducing what an application looks like does not establish how it handles data or business rules. Developers building a replacement or compatible system need to understand what records are stored, how calculations work and which conditions change an outcome. The report argues that an investigation grounded in a program’s behavior can turn those questions into a specification that can be tested.
The source illustrates this with Ledger Reborn, a local accounting application its author says was built after investigating QuickBooks Desktop. According to the report, its profit-and-loss statement, balance sheet and trial balance matched Intuit’s official sample company account by account, to the cent. That is a claim in the source, not an independently verified result in the material provided. If accurate, it shows why tracing specific calculations can matter more than copying interface elements; it does not establish that REA can reproduce other programs or that the recreated app is suitable for production accounting.
The distinction between evidence and model-generated guesses matters for reliability and security as well. An agent that never calls REA may answer from its training or make assumptions based on visible screens. Even when it does use the tool, findings need checking: the tool can provide evidence, but developers must decide whether that evidence supports the conclusion and whether the software’s behavior has been covered adequately.
Products Worth Considering
Logitech MX Master 4 for Mac Wireless Ergonomic Mouse, White Silver
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Brother SP1 Dye-Sublimation Printer
As an affiliate, we earn on qualifying purchases.
How REA Fits Into Agent Work
The report describes REA as an open-source project hosted on GitHub under morluto/rea, with an MIT license. It says version 6.7.0 was released October 18, one day before publication. The source also reports roughly 325,000 stars and 82,000 forks as of October 19, following attention on GitHub’s trending page earlier in the month. Those figures are attributed to the article and represent project activity, not proof of accuracy or adoption in professional development.
The case study discussed in the report examined the QuickBooks Desktop for Mac 2024 trial with REA 6.4.0 and Ghidra. The author says the investigation focused on inventory arithmetic, how account types affect signs in budgets, reconciliation balances and branches used in cash-basis reports. This provides a concrete example of narrowing an inspection to particular rules rather than treating reverse engineering as an open-ended request.
REA runs analysis on the user’s machine, according to the source, but that does not mean all information stays private. Results are passed to the coding agent, and the model provider may receive them under its own data policy. The project also warns that REA is not a sandbox: opening an untrusted program exposes it to analysis tools running with the user’s permissions.
“REA supplied evidence and imperf”
— Thorsten Meyer, describing the Ledger Reborn case study
Limits of the Reported Findings
The source material ends partway through its account of the QuickBooks case study, after the incomplete phrase “REA supplied evidence and imperf”. It therefore does not provide the full case-study explanation, the complete methods or enough detail to evaluate how the reported accounting matches were tested. The comparison with QuickBooks is attributed to the case-study author and is not independently confirmed by the material available here.
It is also unclear from the supplied text how well REA performs across different software types, what errors or blind spots its analysis engines may have, and how much technical expertise is needed to assess findings. The report’s GitHub activity figures do not answer those questions. Its privacy and security implications also depend on what data an investigation exposes to the agent and model provider, and on the software being inspected.
Verify the Connection and Findings
For developers following the guide, the next step is to install REA, review its setup plan, run the prerequisite check and confirm from inside the coding agent that its tools are available. The source recommends beginning with a harmless application the user owns, then testing a narrowly framed question against a concrete example. Before using findings to build or change software, developers should check which statements are directly observed, which are inferred and which remain unknown.
The supplied report does not announce a separate product launch or describe a future milestone beyond the cited REA release. Further detail on the QuickBooks case study, including the full account of its methods and test results, would be needed to assess the claim of matching statements more fully.
Key Questions
What is Reverse Engineer Anything?
REA is an open-source project that the report describes as a local MCP server, command-line tool and agent skill for helping coding agents inspect existing software and data.
Does REA keep all analysis data private?
No such guarantee is established by the source. Analysis runs locally, but results go to the coding agent, and the model provider may receive them under its own data policy.
How should a developer start an investigation?
Pick one behavior and the decision it informs. Ask the agent to explain the relevant stored data, calculation and conditions, and request a concrete example with observed, inferred and unknown claims separated.
How can I tell whether the agent is using REA?
The report recommends checking that the agent can call an REA tool after setup. A prerequisite diagnostic alone does not verify a live connection; an agent answering without calling REA may be relying on guesses rather than an inspection.
Does the QuickBooks example prove REA can recreate accounting software?
No. The source reports that Ledger Reborn matched several statements for an official QuickBooks sample company, but the supplied material does not include the complete case study or independent verification. The example is a reported result, not proof of general performance.
Source: ThorstenMeyerAI.com
