@manateelazycat: Holy crap, WeChat and WorkBuddy have already been cracked by REA Everyone knows that when doing business, WeChat chat r…

X AI KOLs Timeline News
reverse-engineering wechat local-llm open-source agent data-extraction

Summary

作者发现 REA 结合算力舱与本地大模型（如 GLM 5.3 Flash、Qwen 3.8 27B）在一小时内成功逆向 WorkBuddy 读取微信聊天记录的机制，称赞其本地模型不会像在线模型那样拒绝，并开源了相关逆向方案。

Holy crap, WeChat and WorkBuddy have already been cracked by REA Everyone knows that when doing business, WeChat chat records are extremely valuable data, but if we usually try to mess with WeChat, it's easy to get banned. Recently, I discovered that WorkBuddy can actually directly read WeChat chat records So I used REA combined with a computing pod to figure out how WorkBuddy reads them, and I cracked it in just one hour—too damn impressive My actual test results: If it's one computing pod, REA paired with Qwen 3.8 27B is a bit slow. If you want it faster, it's two computing pods with GLM 5.3 Flash I tried it this morning, and it's really too damn impressive! It won't refuse like online models do—it's just this reverse engineering tool REA plus a local large model, truly point-and-shoot accurate, open-source to the world Bosses interested in this reverse engineering setup, drop a 1 in the comments, and I'll give you the materials for free
Original Article
View Cached Full Text

Cached at: 10/13/26, 05:47 PM

Holy crap, WeChat and WorkBuddy have already been cracked by REA

Everyone knows that when doing business, WeChat chat records are extremely valuable data, but if we usually try to mess with WeChat, it’s easy to get banned. Recently, I discovered that WorkBuddy can actually directly read WeChat chat records

So I used REA combined with a computing pod to figure out how WorkBuddy reads them, and I cracked it in just one hour—too damn impressive

My actual test results: If it’s one computing pod, REA paired with Qwen 3.8 27B is a bit slow. If you want it faster, it’s two computing pods with GLM 5.3 Flash

I tried it this morning, and it’s really too damn impressive! It won’t refuse like online models do—it’s just this reverse engineering tool REA plus a local large model, truly point-and-shoot accurate, open-source to the world

Bosses interested in this reverse engineering setup, drop a 1 in the comments, and I’ll give you the materials for free

Andy Stewart (@manateelazycat): GLM 5.3 Flash 本地写代码太爽了，她写代码的时候，我就看着算力舱的GPU利用率发呆，哈哈哈哈

新时代监工 ；）

Similar Articles

@leaf_sanren: https://x.com/leaf_sanren/status/2073069608437764266

X AI KOLs Timeline

The author details how to crack the encryption system of WeChat Mac version through reverse engineering, using Frida dynamic instrumentation to capture the key generated by the system function CCKeyDerivationPBKDF, thereby decrypting 29 encrypted databases. The goal of automatically obtaining group chat records is achieved, and the project is planned to be open-sourced.

@jinchenma_ai: Whoa! Just discovered a tool that can read WeChat chat records – a must-have for private traffic operations! wx-cli enables AI to freely read WeChat messages. 1. Browse messages at will: No need to use WeChat’s clunky search bar; simply enter keywords to search through your entire chat history with lightning-fast speed. 2. Moments excavator: You can directly view friends’...

X AI KOLs Timeline

wx-cli is a local tool for extracting and analyzing WeChat chat history and moments, enabling AI integration without transmitting data to the cloud.

@zjp1997720: [Reverse-engineered WorkBuddy, open-sourced a skill for searching WeChat articles] A while ago, when I dissected WorkBuddy's built-in Deep Research feature, I found that during execution it came with a skill for searching WeChat public account articles, called WeChat Article Se…

X AI KOLs Timeline

The author reverse-engineered WorkBuddy's Deep Research feature, discovered its built-in skill for searching WeChat public account articles (WeChat Article Search), and open-sourced it. This tool implements WeChat public account article search by calling the Sogou search API, returns structured JSON results, requires no API key, and supports one-click installation via npx.