One of the hottest AI developer topics on X over the past day. At 21:55 UTC on 2026-10-23, security account @0x0SojalSec (Md Ismail Šojal) posted that "Claude can no longer Reverse Engineer Anything" and that "Anthropic just quietly killed REA", tying the timing to viral cases of people using Claude to reverse engineer Samsung TV firmware and find data-collection services; the post had about 8,345 likes, 1,055 reposts and 514K views. An earlier post making the same claim by @AbhinavXJ had about 10,477 likes and 640K views. As of this run we found no statement from Anthropic or its staff, and some users replied that it still works for them. These are user reports and opinions on X, not a confirmed policy change.
Key Takeaways
- ✓Main post: @0x0SojalSec's post, about 8,345 likes, 1,055 reposts, 2,545 bookmarks and 514K views when we pulled it
- ✓An earlier post making the same claim: @AbhinavXJ's post (11:16 UTC on 10-23) said "anthropic just killed the REA" and argued this is why local models matter; about 10,477 likes and 640K views
- ✓One trigger was a Samsung TV firmware teardown: @log4j posted that the TV classifies the audio it hears before reporting back, about 6,720 likes; in another post he said his OpenAI account was banned after local reverse engineering work tripped its checks, about 6,454 likes and 729K views
- ✓REA is an open-source reverse engineering toolkit (GitHub morluto/rea), showing about 377K stars in the GitHub API this run; we covered its viral launch on 10-08
- ✓We found no response from Anthropic's official accounts or staff this run; some users replied that REA still works for them and is refused only in specific cases, so the real scope awaits an official word
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and the problem
REA (Reverse Engineer Anything) is an open-source reverse engineering toolkit that lets coding agents such as Claude Code, Codex and Cursor analyze binaries, APKs and Electron apps; the GitHub API showed about 377K stars on the repo this run. Over the past few days, security researcher @log4j and others used AI-assisted reverse engineering on a Samsung TV's firmware and posted that the TV classifies the audio it hears and counts events such as dog barking and baby crying before reporting back; those posts drew thousands to tens of thousands of likes on X. Soon after, several posts with around ten thousand likes claimed Anthropic had "quietly killed" Claude's reverse engineering ability.
How it reportedly works
Users say that when they give Claude a reverse engineering task, the answer is cut off mid-response by a safety classifier or refused outright with a reference to safety rules. @0x0SojalSec's post tied the timing to the viral Samsung TV case, and @AbhinavXJ's post simply said "anthropic just killed the REA". Other users replied that REA still works for them and is refused only in cases like credential scraping. These are user reports; this run found no response from Anthropic's official accounts or staff, so whether this is deliberate and how far it reaches cannot be confirmed yet.
Benchmarks and measured results
This is an X discussion with no benchmark involved, and nothing has been published officially. The numbers we can check are engagement: the main @0x0SojalSec post had about 8,345 likes, 1,055 reposts and 514K views; the @AbhinavXJ post about 10,477 likes and 640K views; and @log4j's post saying his OpenAI account was banned over local reverse engineering about 6,454 likes and 729K views. The main thread of opinion is that local and open-weight models matter more; for example, kache (@yacineMTB) posted that with DeepSeek Flash out it is already too late for such limits, which is a personal view.
What developers can do now
If your workflow relies on Claude for reverse engineering or security analysis, make sure the target is software you own or are authorized to analyze, and watch for updates to Anthropic's usage policy. When a response gets cut off, keep the exact prompt and error so you can tell a classifier false positive from a policy change. Teams that need dependable reverse engineering can evaluate local or open-weight models as a fallback, while checking each provider's terms so they don't trip account checks the way @log4j describes.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.