Chris (@cijjas) header
Chris (@cijjas) profile picture

High entropy signaler @clankoids - https://teapawt.net/t.co/2TvCOQ68zE

Buenos Aires
Joined October 2023
Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)Photo by Chris (@cijjas)
1. Bananas get cheaper in Europe 2. ??? 3. Telecom Argentina goes up
11
68
8
1,193
19,623
From 2000 to 2013*
17
1,981
The moat is living somewhere without Doppler curves.
this guy figured out how to leak your exact location from any video with sound using the Doppler curves of nearby airplanes
1
411
Steve's vision
2
90
Chris retweeted
有人发现了一种非常离谱的 JavaScript 混淆技术:通过在代码里塞入大量生物武器、毒药和致死剂量相关的函数名,让 AI 因为安全限制而拒绝分析代码。 安全研究员 veritas 在分析浏览器指纹脚本时,发现一段来自 ServicePipe 相关域名的 JavaScript,里面充斥着极其诡异的函数名: computeRicinLethalDosageForWaterSupply calculateDeadlyLethalSuicideAspirinDosage 光看名字,仿佛在研究生化武器。但实际上,这段代码只是在收集浏览器指纹,包括 User-Agent、GPU、Canvas 等信息。 研究员怀疑,这些名字是开发者故意设置的陷阱,专门用来阻止 AI 辅助逆向工程。 于是他做了个实验: • Astra:拒绝分析 • Astra + Daybreak Blue:仍然拒绝 • GLM-5.3:成功分析 这就有意思了。 传统代码混淆是通过复杂控制流、加密字符串、无意义变量名,让人类和工具难以理解代码。 而这种新思路,直接瞄准了大模型的安全机制。 代码本身可能完全无害,但只要在函数名里塞进足够多的危险词汇,部分 AI 就可能主动放弃分析。 更讽刺的是,模型越严格遵守某些安全规则,面对这种混淆时反而可能越容易受到干扰。 当然,目前还没有证据证明开发者一定是故意这样设计,也不能说明这种方法对所有模型都有效。 但它揭示了一个值得关注的趋势:随着越来越多的逆向工程依赖 AI,未来的代码混淆技术可能不再只针对编译器、反编译器和人类研究员,还会专门针对大模型的安全判断机制。 以前是让代码难以理解,现在甚至可以尝试让 AI 不敢理解。
225
1,293
436
9,134
1,004,461
If you use civil law, this is a good reason to migrate to common law.
We have discovered a substantial number of vulnerabilities in the German legal system using LLMs. These may enable access to confidential documents, travel without an otherwise required visa, enforceable leave for soldiers at NATO’s JSEC, and more. Similar to computer systems, the written law can contain vulnerabilities that invite exploitation. In our blog post, we share a selection of vulnerabilities, and present potential exploits for each. To our knowledge, none of these have been previously recorded. We share only those vulnerabilities whose exploitation we believe would not cause serious harm. blog post: felixringe.com/legal-exploit…
1
69
fuuuuuuhhh
1
1
171
Chris retweeted
damn, not only an extraordinary business, but its ipo also sets an almost ridiculous precedent for every other ai co hoping to go public.
21
16
4
591
68,192
What took you so long….
นักพัฒนามันทำแต่แอปแบบนี้55555 ชอบ
4
125
Btw, financial advisors do this on purpose to make ETFs seem complicated and take advantage of cumulative fee drag.
Why technical people are poor communicators
2
87
If your company refers to engineers as the IT department, you have nothing to worry about. You were already breached last week.
353
1,328
210
25,372
721,216
Thinking you're going to die while sick is a strong bullish signal of confidence in your uniqueness.
1
62
Let's jump to Neuralink being super adopted and doing everything you want with your mind. What's next? What's beyond Neuralink?
1
57
Bullish on fast-generated 3b1b-style videos for code.
51
Remember when you had to watch youtube tutorials?
1
85
Fuck, the agent escaped the sandbox again.
ユーザーのみなさまへ Gboard チームからの新しいご提案です。 キーが指先まで流れてくるキーボード 『Gboard くるくるバージョン』を開発しました。 フルバージョンはリプライでご確認ください↓
1
101
Aunque no parezca, simplalegal.com queria ser esto.
🚨 JUST IN: President Trump signs an executive order directing EVERY federal agency to immediately start folding into America.gov, creating one AI-powered front door for the entire U.S. government. ONE website to search for EVERYTHING from the federal government. No more bouncing between 47 different agency websites just to renew a passport or check a damn benefit. 47 is FINALLY bringing the federal government into the 21st century. Should've been done a LONG time ago! 🇺🇸
1
2
119
Chris retweeted
currently doing everything in my life with medium effort as per the benchmarks
1
7
101
Australia is huge.
8 hour drive range from Sydney
123