淚笑

- 专注于 AI 应用与 Agent 工程前沿探索
- 深耕安全攻防方向

[淚笑 @le1xia0]

@realfxw I now feel that Cairn's naming really isn't good. I've seen more than half of people misspell this word. It's actually not "carin," but "cairn." https://t.co/uPMoIhsYBE
🤣🤣

[淚笑 @le1xia0]

Anyone interested in the ARTEX and Cairn core architectures is welcome to reach out to me for a chat. I'd be happy to discuss Agent design.

[淚笑 @le1xia0]

I'm the author of Cairn.
Recent reports say ARTEX was used in attacks on South Korean banks, and some people have started linking us.
ARTEX officially listed Cairn as an Architecture Reference in its README:
https://t.co/uPMoIhsYBE
Cairn was never a “hacking tool.”
#Cairn #ARTEX

[淚笑 @le1xia0]

@vysecurity I talked with the author of ARTEX, and he shut down ARTEX because he was worried about legal risk.

[淚笑 @le1xia0]

Whether someone uses it for authorized testing or other purposes is entirely up to the user.

Misuse is a human decision, not a feature of the engine.

[淚笑 @le1xia0]

It’s a general-purpose state-space search engine.
Penetration testing was just the first domain we used to validate the idea.

ARTEX itself is no longer public.
Cairn, the general engine it referenced, is still open.

[淚笑 @le1xia0]

用了一年多的 Claude 账号突然被封了，申诉也没用，还不退款，A/ 😊

@AnthropicAI @claudeai

[淚笑 @le1xia0]

@eyalsela As the author of Cairn (https://t.co/uPMoIhsYBE), I must clarify that Cairn is actually an innovative type of multi-agent architecture and was not designed for hacking attacks. If it is used to carry out harmful hacking attacks, that is not my intention.

[淚笑 @le1xia0]

Cairn goes beyond pentesting — it's a general problem-solving engine for any scenario with clear start, clear goal, and unknown path (CTF, vuln research, complex planning, etc.).

Docker one-click deploy. Stars & contributions welcome!

#Cairn #AIPentest #OpenSource
(4/4)

[淚笑 @le1xia0]

Just open-sourced Cairn — a general-purpose state-space search engine, first validated on autonomous pentesting.
It uses Blackboard Architecture + Stigmergy: agents coordinate only through a shared board, with no fixed roles.
https://t.co/uPMoIhsYBE
@Dinosn @The_Cyber_News
(1/4)

[淚笑 @le1xia0]

Proven in Tencent Cloud Hackathon AI Penetration Testing Challenge (610 teams, 1,345 participants):

→ Only team to AK all 54 challenges
→ 3rd place overall

System first ran completely at 4 AM on race day.

(3/4)

[淚笑 @le1xia0]

@NFTCPS 这里有一个类似的项目：https://t.co/uPMoIhsYBE

[淚笑 @le1xia0]

@QingQ77 刚刚已经公布源代码了，其实不止渗透测试，Cairn 是一个通用的问题探索和解决引擎

[淚笑 @le1xia0]

@chenchengpro 所以为什么觉得 “它不会「扫一眼目录结构猜个大概」” ？

[淚笑 @le1xia0]

@dotey 倒反天罡了兄弟，claude agent sdk 只是包装了一层，这个sdk内置了这个cli … ，你说什么基于这个sdk 写一个 claude code 完全是套娃

[淚笑 @le1xia0]

@tvytlx 真的是不负责任的胡乱吹捧…

[淚笑 @le1xia0]

@sitinme playwright mcp 或者 playwright cli（现在更流行cli） 他们获取网页内容都是 yaml 格式的 无障碍快照，操作和点击也是定位其中的元素

[淚笑 @le1xia0]

@sitinme playwright 不一直都可以吗

[淚笑 @le1xia0]

@mdancho84 I’ve released a new agent framework: https://t.co/bEq64qmN37.

Formalizes human intent and executes it directly in Python.
Replaces discrete function calling with a more expressive DSL.
Separates data from instructions to address LLM context limits.