An open-source MCP tool lets coding agents crack open compiled binaries with no source code, and Rob Shocks tests it against his own encrypted app to see if the hype holds up.
REA doesn't invent new reverse-engineering power, it removes the friction between a coding agent and professional tools like Ghidra and IDA Pro, so any app you ship as a compiled binary should now be assumed readable by someone else's AI agent.
REA is an open-source MCP server and CLI that lets a coding agent inspect a compiled app it has no source code for, trace how a feature works, and hand back pseudocode and evidence instead of the original source. It isn't a new decompiler, it wraps proven tools like Ghidra, Hopper, and IDA Pro and removes the expertise barrier that used to protect closed-source software. Rob tests it against his own stripped binary and watches it recover the unlock logic and generate working code without ever seeing the source. The takeaway: the time and skill reverse engineering used to require was the real moat, not the compiled code itself, so what you ship now needs a different kind of protection, and running REA against anything you don't own should happen in a sandbox.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →
Rob opens with the hook: REA lets a coding agent reverse-engineer apps it has no source for, and it's pulling thousands of GitHub stars a day.
You show your agent a feature you want in your own product, it investigates the target app and explains how that feature is built, without needing the app's source code.
REA runs as an MCP server and CLI on native binaries across Mac, Windows, and Linux, plus .NET, APKs, firmware, and smart contracts. It isn't a new decompiler, it wraps proven tools like Ghidra, Hopper, and IDA Pro and gives an agent plain-text tools on top. Setup is one command that asks before touching anything critical.
A sponsored segment for TestSprite, an AI testing agent that writes and runs end-to-end tests against a real running app and hands failures back to the coding agent to fix.
Rob builds his own password-vault app with every function name stripped, to run a fair test with a real answer key. Pointed at the binary with Ghidra, REA pulls back 242+ unlabeled functions, decompiles them, and finds the unlock rule a handful of calls in.
REA didn't copy the source, it understood the logic and wrote a brand-new code generator the app never had. For years the slowness and cost of reverse engineering was the real protection for shipped software, and that friction is now gone.
Reverse engineering for interoperability is generally protected, but many apps' terms of service ban it outright. REA isn't a sandbox, so it runs with your real permissions, and the project is changing fast enough to need constant updates.
Rob lists his own use cases: learning how well-built apps solve problems, auditing Electron apps already on his machine, and recovering logic from his own old builds.
A soft pitch for Switch Dimension's waitlist and a subscribe ask.
REA doesn't add new reverse-engineering power, it strips away the friction between a coding agent and professional tools, so any compiled app you ship should now be treated as readable.
“That friction of reverse engineering was the protection, and REA basically deletes all that friction.”
“Code was never really the moat in the first place.”
“Your moat really is in your data, your users, your cloud security, how fast you ship, your taste.”
“It didn't copy the code. It understood the logic and then wrote something completely new.”
“There could be hidden malicious instructions for your AI. Run this stuff in a sandbox.”
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
A new open-source tool lets your coding agent crack open an app it has never seen the source for, explain how a feature works, and build you a version of it. It's called REA, and it's one of the fastest-rising repos on GitHub right now, pulling seven or eight thousand stars in a single day.
The loop REA follows: your agent asks a question, REA inspects and traces the target binary, surfaces the evidence it found, and your agent uses what it learned to explain or implement the feature.
“If you want to go deeper on building with agents, MCPs and things like this, that's exactly what we do in Switch Dimension, the course and community. It's closed at the moment, the waitlist is open.”
Soft pitch folded into the 'what I'd use it for' wrap-up rather than a hard sell, followed immediately by a plain subscribe ask.
00:00
00:12
00:14
00:22
00:32
00:40
00:46
00:51
00:59
01:05
01:1301:20
01:22
01:32
01:38
01:49
01:53
01:58
02:06
02:12
02:20
02:2702:31
02:44
02:50
02:53
03:03
03:0803:15
03:24
03:27
03:35
03:45
03:48
03:56
04:05
04:07
04:17
04:24
04:31
04:40
04:44
04:50
04:55
05:0705:08
05:19
05:25
05:32
05:39
05:46
05:51
06:00
06:07
06:12
06:18
06:27
06:3706:40
06:45
06:56
07:03
07:08
07:14
07:22
07:27
07:3807:46
07:50
07:55
08:03
08:0808:18
08:25
08:31
08:38
08:42
08:49
08:5709:02Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
16:21Rob Shocks breaks down Anthropic's new AI-native SDLC playbook, the intent.md to plan.md artifact chain meant to keep agents and humans in sync from planning through autonomous maintenance.
September 1st
12:57A walkthrough of Lauren Tan's pstack: 21 engineering principles, 22 playbooks, and 24 skills that turn a coding agent from a slop machine into a verification-obsessed engineer.
September 8th
10:36TypeSpace's new model can't write you a sentence, but it can make a decision 200 times cheaper than the LLM you're already paying for.
September 17th
09:04How a 9-minute validation interview against your own idea can save you months of building the wrong thing.
June 5th
15:32Rob Shocks audits every major Claude Code skill library and collapses them into one five-step loop you can run from memory.
May 18th
Video of the Day14:07Rob Shocks unpacks Andrej Karpathy's AI Ascent talk into four frameworks every builder needs in their head, then runs his own side-project folder through the test and kills three apps live.
May 1st