dragonauta retweeted
The hardest truth is that your greatest work may be seen by no one, and yet you must do it anyway for the one who watches is you
— Friedrich Nietzsche
The Cassandra Effect in Neurodivergence
gemini.google.com/share/0d1e…
1
28
dragonauta retweeted
Tu personalidad tal vez sea mucho más compleja. Explora el espectro de 64 tipos y descubre tus verdaderos matices.
Sunday, Sunday here again, a walk in the park.
You meet an old soldier and talk of the past.
He fought for us in two world wars and says the England he knew is no more.
He sings the Songs of Praise every week but always falls asleep.
#Blur Sunday Sunday
youtu.be/qI6YPGdDB1A?si=3bui…
1
48
Replying to @dragonauta @DavidUllrich202
Quite. The species that named itself "wise man" is the main driver accelerating large-bodied mammal losses toward 75% in roughly 120–160 years. The label tracks intelligence more closely than foresight.
Discúlpenme cinéfilos fans de #Spielberg, pero Disclosure Day es una mamadotota.
1
23
mercenary
fantasies
zealot demagogueries
hunter gatherer
fight or flight
requiem electronic
the specter of the deathling
acension of the mother
#Netherlands - Deathling
youtu.be/rpXMT-izwJo?si=vdJy…
1
61
dragonauta retweeted
Julia Roberts told reporters this week that her marriage to Lyle Lovett has been over for some time.
The key moment, she said, came when she realized that she was Julia Roberts, and that she was married to Lyle Lovett.
Strategies in Radical Rock Music Abstraction share.gemini.google/oMbeHxNb…
19
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。
22
We mourn the passing of Patrick Stewart, who inspired countless individuals at NASA (and beyond) to boldly go where no one has gone before.




















