The Solo Economy

Software Can Be Read Now. The Fit Cannot.

软件现在读得懂了。合不合你,读不出来。

In plain English

A free project on GitHub called REA, short for Reverse Engineer Anything, went viral this month. It lets an AI agent look at an app, work out how one feature is built without ever seeing the app's code, and then build a similar version for you. People are saying software is dead. I don't think so. A software subscription always charged you for two things mixed together: how the feature works, and whether it fits and keeps running inside your business. The first half just became easy to read. The second half was never inside the app.

这个月 GitHub 上有一个免费项目爆红,叫 REA,全名 Reverse Engineer Anything。它让 AI 打开一个 app,不用看源代码,就能弄懂某个功能是怎么做的,再帮你做一个类似的。很多人在说「软件死了」。我不这样看。你付的软件月费,一直是两样东西混在一起收的:一个是这个功能怎么运作,一个是它合不合你、能不能天天在你的生意里跑。第一样,现在很容易被读懂了。第二样,从来就不在 app 里面。
TL;DR

REA connects an AI agent to the tools experts use to take software apart. Anyone with Claude Code or a similar agent can now ask "how does this feature work?" and get an answer with evidence. That puts pressure on software whose only defence was "you can't build this yourself." It does not touch the half you can't read from the outside: your data, your workflow, and someone keeping it running. That half is where the Forward Deployed Engineer works, and where a One-Person Company wins.

REA 把 AI 跟专家拆软件用的工具接在一起。现在有 Claude Code 这类工具的人,都可以问「这个功能怎么做的」,然后拿到有证据的答案。只靠「你自己做不出来」撑着的软件,压力会很大。但从外面读不到的那一半,它碰不到:你的资料、你的流程、还有人帮你一直跑下去。那一半,就是 FDE 的工作,也是一人公司赢的地方。

What actually happened

The project is called REA. It sits on GitHub under the account morluto, and it is free under the MIT licence.

The growth was fast. 36Kr reported nearly 70,000 stars in two days, and that it topped GitHub's trending list. When the repo page was checked on 13 October it showed 158,800 stars and 36,900 forks.

The README describes it in one line: "One MCP for reverse engineering across binaries, applications, and runtime behavior." In plain words, it is a plug. On one side is your AI agent. On the other side are the specialist tools people use to take software apart, like Ghidra, IDA, JADX and others.

It reads a wide range of things: desktop apps, phone apps, websites, games, even firmware. And it works with the agents many of us already use every day, including Claude Code, Codex, Cursor and Gemini CLI.

Inspect

The agent looks inside an app it has no source code for.

Explain

It tells you how one feature works, and shows the evidence.

Rebuild

It builds a similar version for your own project.

Its own showcase is honest about scale. One example rebuilt the sound logic of the old game DX-Ball and passed 3,205 test cases. Another traced how Notion's desktop app moves copied text between its parts. These are features, not whole companies.

"Software is dead." I read it differently.

The line going around on X was "Software is dead, long live software!" 36Kr went further and predicted that small code-level advantages will stop protecting anyone.

I understand the feeling. But I think two things are being mixed together here, and it helps to pull them apart.

When you pay a monthly fee for software, you are paying for two halves.

Half one · now readable

The feature

  • How the button works
  • The rule behind the calculation
  • The format the app sends to its server

Lives inside the app file. REA can read it.

Half two · not in the file

The fit

  • Your data, and years of it
  • How your team actually works
  • Someone fixing it at 9pm when it breaks

Lives in your business. Nothing can read it from outside.

REA makes half one cheap. Half two was never inside the app, so there is nothing to read.

A security writer, Rafael Roman, made the same point about banking apps. Take one apart and you learn how it asks for things. You do not get the power to move money, because the bank's server makes that call. In his words: "The app asked; the server decided."

Software is not dead. One half of its price just got cheaper.

Weiss Ang

What this means for SaaS

The pressure will be uneven. Harvard Business Review already used that word for AI and SaaS back in May.

  • Software that is mostly one clever feature gets the most pressure. If the only reason to pay was "you can't build this yourself," that reason is getting weaker every month.
  • Software that holds your data, connects many people, or makes the real decisions on its own server keeps its ground. Those things don't sit in the app file.
  • Software with real care behind it keeps its ground too. Updates, support, security. That is work, not code.

So the question every software company now has to answer is simple. When a customer pays you, which half are they paying for?

What this means if you work like an FDE

In August I wrote that no small team defends a technical advantage for long, and that the moat is execution and what you learn inside one industry. REA pushes that same rule onto software companies.

For a Forward Deployed Engineer, the client's question changes.

It used to be "which tool should we buy?" Now it is closer to "what should we build for us, and is it worth owning?"

Knowing how a feature works is becoming cheap. Knowing which feature this client actually needs, which ones to leave out, and making it run inside their messy, real systems is not. That second skill is the job. It always was. Now it is the obvious part of the job.

What this means for a One-Person Company

For a one-person business this is good news, with a catch.

The good news: you no longer have to rent ten subscriptions to get ten small features. With an agent and a tool like REA, you can learn how a feature works and build your own version, shaped to how you actually work.

The catch: owning means carrying. A subscription you pay for, someone else fixes. A tool you build, you fix. Building is now cheap. Running is still work.

So I would not cancel everything. I would sort.

The line I hold

In my own system there is a written rule for any outside codebase I study: take the pattern, never the code. I learn how something is done, then build my own way. If there is no licence, nothing from it ships.

REA's README draws a similar line. It says the project is for "lawful research, education, interoperability, software preservation," on software you own or are authorised to examine. On Hacker News, a founder building Photoshop-like and Word-like apps said they keep their work "100% clean room" and warned that using REA on a high-profile product "is likely to result in lawsuits."

Rules on this differ by country and by licence agreement. If you plan to sell something built from what you learned, talk to a lawyer first.

Start this week, no tool needed

You don't need REA to use the idea.

  • List every software subscription you pay for. Your bank statement is enough.
  • Next to each one, write one word: feature or fit. Are you paying for one small thing it does, or for it running your business every day?
  • Keep the fit ones. They are doing work you would otherwise carry.
  • Look hard at the feature ones. Those are the ones you may be able to build, simplify or drop.

That one sort tells you more about the future of software in your own business than any headline.

If you want the bigger picture, start with what a One-Person Company actually is. The whole system I use is free in my community.

Weiss Ang
Who am I

I'm Weiss Ang. One person, plus an AI department, is a company.

I ran mine first, from Sabah, Malaysia. Now I help people in their own industry build a line that keeps running by itself. I have trained more than 1,300 people to use AI, and I run the One-Person Company community, where the whole system is free.

我是洪家徽。一个人,加 AI 部门,就是一家公司。我先在沙巴把自己的跑通了,现在帮你那一行,搭一条会自己跑的线。我教过一千三百多人用 AI,也在经营 One-Person Company 社群,整套系统免费。

FAQ

What is REA?
REA stands for Reverse Engineer Anything. It is an open-source project on GitHub, published under the MIT licence by the account morluto. It connects an AI agent such as Claude Code, Codex, Cursor or Gemini CLI to reverse-engineering tools, so the agent can inspect an app it has no source code for, explain how a feature works, show the evidence, and build a similar version for your own project.
Is SaaS dead because of tools like REA?
No. What changes is the price of one half of it. A SaaS subscription has always paid for two things mixed together: the feature, meaning how it works, and the fit, meaning it runs every day inside your business with your data and someone looking after it. REA makes the feature much easier to read. The fit was never inside the app file, so it cannot be read out of it.
Is it legal to use REA on someone else's software?
REA's own README says it is for lawful research, education, interoperability and software preservation, on software you own or are authorised to examine. Rules on reverse engineering differ by country and by licence agreement. If you plan to build a product from what you learn, get legal advice first and take the idea, not the code.
What does this change for a Forward Deployed Engineer?
The client's question moves from which tool to buy to what should be built for this business, and whether it is worth owning. Knowing how a feature works becomes cheap. Knowing which feature this client actually needs, and making it run inside their real systems, stays scarce. That is the FDE's job.
Should a one-person business stop paying for software?
Not across the board. Building a version of a feature is now cheaper, but running it is not free: someone has to fix it, update it and keep your data safe. Keep paying for the subscriptions where you are buying the fit. Review the ones where you only use one small feature.

Sources

Build your One-Person Company with me.

The whole system and the course are free inside my community. Put your name on the waiting list.

Join the waiting list