<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: AIshihzong</title>
    <description>The latest articles on DEV Community by AIshihzong (@aishihzong).</description>
    <link>https://dev.to/aishihzong</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4237182%2Fb8201641-66e2-4ea8-904a-a5b7639f9df1.png</url>
      <title>DEV Community: AIshihzong</title>
      <link>https://dev.to/aishihzong</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/aishihzong"/>
    <language>en</language>
    <item>
      <title>We checked 39 viral "built with REA" demos. Here's what they actually used.</title>
      <dc:creator>AIshihzong</dc:creator>
      <pubDate>Thu, 12 Nov 2026 15:41:00 +0000</pubDate>
      <link>https://dev.to/aishihzong/we-checked-39-viral-built-with-rea-demos-heres-what-they-actually-used-4hgf</link>
      <guid>https://dev.to/aishihzong/we-checked-39-viral-built-with-rea-demos-heres-what-they-actually-used-4hgf</guid>
      <description>&lt;p&gt;When &lt;a href="https://github.com/morluto/rea" rel="noopener noreferrer"&gt;REA (Reverse Engineer Anything)&lt;/a&gt; took off in October, "I reverse-engineered X with AI" posts got attributed to it almost by default. Press did it too: GTA V on iPhone, Adobe clones and game ports were all filed under REA.&lt;/p&gt;

&lt;p&gt;I maintain &lt;a href="https://github.com/AIshihzong/awesome-reverse-engineer-anything" rel="noopener noreferrer"&gt;Awesome REA&lt;/a&gt;, a bilingual list of what AI agents reverse-engineered and rebuilt. Since the label had drifted so far, I went case by case and checked each claim against the project's own README, its agent instructions (AGENTS.md, CLAUDE.md), the author's post, or REA's official showcase.&lt;/p&gt;

&lt;h2&gt;
  
  
  The numbers (39 cases, re-checked 2026-11-10)
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;26 confirmed:&lt;/strong&gt; a primary source says REA was used.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;3 partial:&lt;/strong&gt; REA was used for one part only. For example, the TH04 showcase used REA for one inspection, while the reconstruction itself builds on ReC98.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;4 not verified:&lt;/strong&gt; two are claimed only in the author's post, and two have no source naming any tool.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;6 not REA:&lt;/strong&gt; mostly Claude Code or Codex driving Ghidra or IDA over MCP, or plain manual reverse engineering.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The most-shared cases skew the other way. Of the six with the biggest reach, only &lt;strong&gt;DX-Ball&lt;/strong&gt; says in its own repo that it used REA. Two more say so only in the author's post. The other three, including &lt;a href="https://github.com/rehan-remade/universal-modder" rel="noopener noreferrer"&gt;Universal Modder&lt;/a&gt; (about 14k likes) and a &lt;a href="https://github.com/GLinnik21/plx-native" rel="noopener noreferrer"&gt;native Plex client for LG TVs&lt;/a&gt;, used other tools or are unverified.&lt;/p&gt;

&lt;h2&gt;
  
  
  Does REA help?
&lt;/h2&gt;

&lt;p&gt;That is a separate question. The two third-party controlled comparisons I know of:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://github.com/pablodiazjorge/rea-lab" rel="noopener noreferrer"&gt;rea-lab&lt;/a&gt;: five preregistered studies, 482 runs with Claude Code. None found a significant gain in solve rate from REA, and REA runs took about twice the time and cost.&lt;/li&gt;
&lt;li&gt;
&lt;a href="https://github.com/aicodeking/leanrea" rel="noopener noreferrer"&gt;leanrea&lt;/a&gt;: 36 runs against REA 6.5.0. A plain-CLI skill scored 74/75 against REA's 75/75, with 38% less wall time and 35% less cost.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Both are small, single-model studies, so treat them as hints, not verdicts. What they do suggest is that the agent and the underlying tools (Ghidra, IDA, JADX) do most of the work, and the orchestration layer is the smaller part.&lt;/p&gt;

&lt;h2&gt;
  
  
  If you want to try it yourself
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;The official package is &lt;code&gt;rea-agents&lt;/code&gt; on npm. Several lookalike repos copy REA's README or ask you to download a zip. They are not official releases.&lt;/li&gt;
&lt;li&gt;If you just want something you can use today, the &lt;a href="https://github.com/AIshihzong/awesome-reverse-engineer-anything/blob/main/docs/by-app.md" rel="noopener noreferrer"&gt;by-app index&lt;/a&gt; lists open-source rebuilds by the app they replace (OBS, Logitech G HUB, Plex on LG TVs, PowerToys and more), with license and status.&lt;/li&gt;
&lt;li&gt;Reverse engineering is governed by local law, licenses and terms of service. Only analyse software you are allowed to.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  How the checks work
&lt;/h2&gt;

&lt;p&gt;Tags reflect what the cited sources say. I haven't reproduced the projects. A validator runs on every pull request to keep the English and Chinese lists in sync, and new cases ship in a &lt;a href="https://github.com/AIshihzong/awesome-reverse-engineer-anything/releases" rel="noopener noreferrer"&gt;weekly digest&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Found a case I got wrong, or one that's missing? There's an &lt;a href="https://github.com/AIshihzong/awesome-reverse-engineer-anything/issues/new?template=case.yml" rel="noopener noreferrer"&gt;issue form&lt;/a&gt;. And if the list is useful, a ⭐ on &lt;a href="https://github.com/AIshihzong/awesome-reverse-engineer-anything" rel="noopener noreferrer"&gt;the repo&lt;/a&gt; helps other people find it.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>reverseengineering</category>
      <category>opensource</category>
      <category>claude</category>
    </item>
  </channel>
</rss>
