Eric Denman 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。
Eric Denman retweeted
The rewrite of the 𝕏 Android app was beneficial not only from a technical perspective.
Replying to @elonmusk
Rewriting the 𝕏 Android app was also a big deal. It was really bad for such a long time.
Eric Denman retweeted
Ranks amongst the top technically challenging project that I worked on
Replying to @elonmusk
Rewriting the 𝕏 Android app was also a big deal. It was really bad for such a long time.
Eric Denman retweeted
Hi X Android folks!
If you're on version 12.20+, you're now running the latest iteration of our new Android app. For context: for the last 8 months we were shipping two apps in a single binary, and that introduced some perf overhead. But shipping two apps in one was needed to make the switch to the new experience as smooth as possible for everyone.
With 12.20+ that overhead is gone. We ship a single app, built entirely on the new shiny kotlin + compose codebase.
This resulted in:
• 33% smaller app binary size
• 69% faster app startup
• Got rid of 3.15M lines of code from the OG app
What's next: we hear you on performance. It's our main focus right now - scroll smoothness, battery, networking and memory - and there's more coming.
Thanks for the patience and the feedback along the way - keep it coming!
Eric Denman retweeted
Exhibit 100 for why we still need to read code.
Good moment today to remind everyone to stay frosty with AI. I was working on making a server stoppable via the CLI, and Codex decided the best path forward was to implement an unauthenticated network API to every server to shut itself down. Excellent work.
I responded with "I don't know about that." (I usually give more direct feedback but this was so absurd) and it thought for like 20 seconds and responded "Server stopping should not be part of an unauthenticated network API." Yes. lol.
By the way, the result I was looking for was that we automatically start a locally discovery Unix-socket bound server so I wanted it to discover that (using prior mechanisms already made) and gracefully stop that.
The driver is absolutely accountable here (me). My lazy prompting led AI to think I wanted to be able to stop ANY configured server I was talking to. Shame on me for thinking it was obvious here. But, I also reviewed it and found this idiocy, so, that's where good AI drivers come in. Also me.
Eric Denman retweeted
New X Android app: 100% Kotlin.
X Chat: Kotlin Multiplatform across Android, iOS, and Web.
E2E encryption, storage, sync, business logic: written once, shipped everywhere 🔐
💬 Which logic would you never want to write three times?
Eric Denman retweeted
Replying to @GergelyOrosz
2.5 years!
I still remember the first commit and the prototype we did to validate if this is even feasible!
The hard part wasn't the UI. It's the analytics, observability, device fragmentation and data persistence.
Eric Denman retweeted
Today we're announcing the completion of one of the largest engineering projects in the company's history:
We rebuilt the X Android app from scratch
It's faster, smoother and more reliable. But most of all: it will enable us to build new features at lightning speed.
Eric Denman retweeted
Messi clearly fouls Spence in the buildup to the second goal, but because it’s not against Argentina the VAR will never go back and check
This media is unavailable
Replying to @YashRMFC
The biggest is the foul of Messi on Spence for the winner, same as Lisandro disallowed goal vs Egypt
Readers added context they thought people might want to know
The supposed contact that Messi had on Spencer would've been his right leg. However, Spencer shows that his left leg is in pain due to an overextension toward the ball instead of the supposed contact from Messi.
The interaction can seen better here.
x.com/Santiag9082817…
Eric Denman retweeted
A tourist was seriously injured after a bison tossed them about 8 feet into the air in Yellowstone National Park. The attack was captured on video by photographer Mike Macleod.
Eric Denman retweeted
An announcement about Retrofit, OkHttp, Okio, and SQL Delight: jakewharton.com/the-lysine-c… – at Ireland
Eric Denman retweeted
An update to the greatest story of all time.
As the #Knicks rallied with a game-winning 2-pointer, NYers held their number 1! At the final whistle & for the next 7 mins, NYC saw a water‑use spike equal to 368,697 🚽 flushes! As the Knicks surged, we fought the urge… because we knew the team was on the verge… OF HISTORY!





























