this is getting f*cking dangerous...
REA creator just dropped 20 GitHub repos to build a f*cking reverse engineering studio:
[01] analysis core
> REA: one MCP that plugs into everything https://t.co/xfRnXmvkwz
> Ghidra: decompiler, free https://t.co/8FLvEkKEze…
> GhidraMCP: lets your agent drive Ghidra https://t.co/Vycy2WICJI…
> ida-pro-mcp: wires your agent into IDA Pro https://t.co/u784y7f6a2…
[02] mobile & apps
> jadx: turns Android APKs back into Java https://t.co/7WHPMvVyUt
> Apktool: pull APKs apart and put them back together https://t.co/zPSDUQVR0m…
> Frida: hook into apps while they run https://t.co/Fc6UHX86Rd
> ILSpy: decompiles .NET https://t.co/Xa5logyrsr…
[03] binary & low-level
> radare2: RE framework that lives in your terminal https://t.co/Q43Dy8oWAP…
> x64dbg: debugger for Windows https://t.co/0i4IvEurbB
> ImHex: hex editor with its own pattern language https://t.co/dVvztlh3QY
> angr: symbolic execution engine https://t.co/sjqE0oOUqE
[04] network & runtime
> mitmproxy: sits in the middle and reads the traffic https://t.co/0KETBkaz1M…
> dnSpy: debug and edit .NET apps https://t.co/tMxhSh02H1
> rizin: the radare2 fork https://t.co/q5FefKTBe3
> raddebugger: the debugger out of Epic Games https://t.co/VdCCwvfZlB…
[05] rebuild & verify
> awesome-ai-reverse-engineering: 122 repos, each one graded https://t.co/IiMIA5oz4r…
> awesome-reverse-engineering: curated by the guy who built REA https://t.co/UE6Kax4dsE…
> REA workflows: ready-made investigation prompts, ship with REA
> clean-room loop: SPEC.md, oracles and checks until it matches
how it runs: binary -> REA -> SPEC.md -> rebuild -> oracle -> retry
save this, point it at any app and watch it explain itself.
Nasdaq invented the electronic stock market. Today, 135 exchanges around the world rely on our technology.
Learn how we bring listed companies and investors closer together through greater transparency, actionable data, and deep liquidity.
The only red mark in last night’s run was the best line in the log.
Six Opus 5.5 agents, 9 hours, and the builder tried to deploy to 5 regions when scope said 1.
That’s not a failed step.
It’s the reason anyone could sleep through the run at all.
Most overnight setups are built to finish.
This one was built to stop at a line it was handed before bed.
What the 9 hours actually looked like:
→ scout read 3,800 files and flagged a stale base before a single branch moved
→ the lead planned 40 rebases across 6 lanes and asked zero questions, because scope was already on paper
→ builder hit 9 conflicts, resolved 9, then held at 1 region instead of shipping to 5
→ CI went green, scout still caught 3 race conditions behind it, and the gate re-ran
→ 12 merged, 28 parked in the queue for a human instead of pushed through
Medium effort, $4 in / $20 out per 1M, $18.60 for the night.
1,240 past runs loaded, nothing lost, review set for 08:00.
Zero prompts was never the flex.
The agent you can leave alone isn’t the one that never hits a wall.
It’s the one that brakes at the wall without calling you.
Every lane and every step of that night is in the recording 👇