joker.icp retweeted
Property tax is the only system where a fully paid-off asset can be legally stolen from you by the government 🤡🌎
joker.icp retweeted
Codex team might have actually fixed the "model at capacity" errors that have been pestering you.
Replying to @gandreou007
"Claude is now running on NEAR AI Cloud" in the same way my toaster is running on Amazon because I ordered it there. Claude runs on Anthropic's servers.
NEAR AI docs: third-party models are "proxied to their providers."
It's a reseller with a token.
Underrated AI chain -> overrated API key.
joker.icp retweeted
I wish I had kept my 1,700 BTC @ $0.06 instead of selling them at $0.30, now that they're $8.00! #bitcoin
Replying to @thsottiaux
When the model reaches capacity, or when it says the model is at capacity, it is crazy stuff that it just stops working. For example I will start work overnight and then I'll wake up in the morning and it has done three turns. Then it just says, "Model is at capacity right now," and it just stops and doesn't retry ever. That is something you could fix with a loop and an if statement and it would make Codex able to be way more trustable for doing long work
joker.icp retweeted
This is the best fu*king interview process.
So a guy I know recently interviewed at a startup.
The interviewer gave him 30 minutes to build a mini AI agent and allowed him to use Claude Code, ChatGPT, and pretty much any AI tool he wanted.
The interviewer left the room while he worked.
After 30 minutes, he came back, asked him to close his laptop, and started asking him why he chose that tech stack, which database he used, why he made those decisions, and how the whole thing worked end to end.
My guy did not have a fu*king clue.
Replying to @zacodil
“The bug was in SGLang, not NEAR AI code.”
Well, thank god nobody touched NEAR code aka the docker-compose.yml.
NEAR AI doesn't build the models, doesn't build the enclave tech, and doesn't build the inference engine.
Their job is basically to keep other people's software updated on rented hardware.
SGLang had three critical 9.8 RCE vulnerabilities disclosed this year. Public PoC since April ... then someone exploited it on November 8.🤣
That's not some sophisticated supply chain attack. That's failing to patch software.
And here's the best part.
Remember NEAR AI big "0 operator access, by design" claim?
Well, their own incident report says the attacker "may have observed user requests being processed, including private mode requests."
So the enclave kept NEAR AI out, but apparently not the attacker. Working as designed, I guess.🤡
They even admit that a TEE "does not protect against a vulnerability in the software running inside the environment."
Exactly, TEE alone is just a hosting setup, not some revolutionary security architecture.
The damage so far?
❌Around 100 users emails, names and balances exposed.
❌18 agents chat logs and configurations.
❌Two customers API keys sitting in plaintext logs.
❌Private Chat taken offline.
❌No CVE number given for the incident.
And all this just 10 days after launching a cybersecurity product. You really can't make this stuff up.
Now pointing out that xAI, Cursor, Oracle, Google and AWS also use SGLang.
Sure, but there is one small difference.
Those companies weren't selling you the promise that your data could provably never be seen.
NEAR AI was.
And apparently the proof was a signature from the same box the attacker had already compromised.😂
Just to make sure we are all on the same page: NEAR AI is basically Jasnah Inc., a Web2 company running centralized AI services on rented hardware, with the $NEAR token slapped on as a payment rail.
Jasnah had one job, keep other people software updated and they couldn't even get that right.🤡
joker.icp retweeted
⚠️ Possible Codex reset by Nov 13, 08:00 UTC
codex-resets.com/
nitter.cf/thsottiaux/status/2120…
Replying to @thsottiaux
Thursday is feeling codexy to me. Perhaps also a good time for a banked reset.
joker.icp retweeted
We ran Fable 5.2 through 100 unsaturated coding and engineering environments spanning spatial reasoning, planning, games, circuit design, and more.
As with all Fable releases, it was an extremely expensive benchmark to run. The biggest winner, however, was OpenAI. Astra, GPT 6.1-Sol, and even GPT 6.1-Terra came up with better solutions to challenging, open-ended technical problems across financial markets, engineering design, and unsaturated coding problems. It's fairly obvious to us why the lion's share of mathematical breakthroughs are coming out of OpenAI.
For what it's worth, Anthropic is all-in on training a good coding assistant (which they are succeeding at), not releasing the smartest model.
Some things we noticed about Fable 5.2:
• It stopped early in our agentic coding runs to ask for human help or clarification "I want you to decide on them before I tune any further" (there is no human to interact with and no reason to expect there would be from the prompt -- we've never seen another model attempt this)
• It became confused in our harness (not Claude Code) and spent time trying to understand past tool calls. We have not seen this behavior from other OpenAI or Anthropic models and it hurt its agentic coding scores.
• It sometimes used all of its reasoning budget and exceeded its token limit
• Several content filter failures (does not affect its score)
From our testing, there is no reason to use Fable 5.2 for any purpose, regardless of budget. Use Opus 5.5 for coding, and Astra or Sol for code reviews and challenging, unsaturated problems in engineering, math, and science.
UN DESARROLLADOR ACABA DE HACER LO QUE GOOGLE LLEVA AÑOS IGNORANDO
Creo un navegador en Rust pensado especificamente para automatizar tareas, web scraping y agentes de IA
> Consume 30MB de RAM
> Las paginas cargan en 85ms
> Bloquea +3.500 trackers automaticamente
> Evita anuncios, analitica y scripts de tracking
> Tiene renderizado nativo. Sin Chromium
Se llama Obscura
Y tiene algo que Chrome jamas va a tener
Cada sesion genera una huella digital diferente. GPU, canvas, audio, bateria... todo randomizado
Los detectores no pueden atraparlo porque se comporta exactamente como Chrome real
Es un reemplazo directo de Puppeteer y Playwright
Sin Nodejs. Sin dependencias. Un solo binario
Tiene MCP y puede sacar screenshots, PDFs y screencast nativos
Es 100% gratis y open source
REPOO 👇
joker.icp retweeted
ELON MUSK: “A conta de luz da IA é alta demais? Então, basta construir um data center no espaço”
“Em vez de economizar energia na Terra, a ideia de se mudar para o espaço”
REPÓRTER: “A IA consome uma quantidade absurda de eletricidade, né?”
ELON MUSK: “É verdade.”
REPÓRTER: “E isso vai aumentar ainda mais no futuro?”
ELON MUSK: “Se a IA evoluir, a demanda por energia também vai crescer, provavelmente.”
REPÓRTER: “Então a conta de luz vai ser um problemão!”
ELON MUSK: “Por isso vamos usar o espaço.”
REPÓRTER: “.... Hein?”
ELON MUSK: “Vamos construir um data center para IA no espaço.”
REPÓRTER: “Peraí, a gente tava falando de conta de luz, cara”
ELON MUSK: “E é exatamente isso que eu tô respondendo.”
REPÓRTER: “Por que raios no espaço, mano?!”
ELON MUSK: “Porque dá pra usar energia solar.”
REPÓRTER: “Energia solar a gente já usa aqui na Terra, né?”
ELON MUSK: “Mas na Terra tem noite, não tem?”
REPÓRTER: “É, tem sim.”
ELON MUSK: “E a geração varia por causa do clima.”
REPÓRTER: “Hmm, entendi.”
ELON MUSK: “No espaço, ajustando a órbita, dá pra captar luz solar por muito mais tempo.”
REPÓRTER: “Então você vai colocar painéis solares no espaço?”
ELON MUSK: “Exatamente isso.”
REPÓRTER: “Mas os computadores da IA não esquentam pra caramba?”
ELON MUSK: “Claro que geram calor.”
REPÓRTER: “No espaço não tem ar-condicionado, né?”
ELON MUSK: “A gente usa dispositivos de dissipação de calor.”
REPÓRTER: “Dispositivos de dissipação?”
ELON MUSK: “Os painéis solares ficam virados pro Sol.”
REPÓRTER: “Tá.”
ELON MUSK: “Do lado oposto, colocamos equipamentos que liberam o calor pro vácuo do espaço.”
REPÓRTER: “Entendi!”
ELON MUSK: “É a radiação do calor pro espaço sideral.”
REPÓRTER: “Então o resfriamento é de graça!?”
ELON MUSK: “Não, ainda precisa de equipamentos.”
REPÓRTER: “Que é que é isso?”
ELON MUSK: “Mas dá pra criar um sistema bem mais eficiente.”
REPÓRTER: “E aí, pra quando isso rola?”
ELON MUSK: “Em 2 ou 3 anos, acho que o espaço vai ser o lugar mais barato pra rodar IA.”
REPÓRTER: “... 2 ou 3 anos!?”
ELON: “É o que eu prevejo.”
REPÓRTER: “Rápido demais, cara”
ELON MUSK: “O progresso da tecnologia é acelerado.”
REPÓRTER: “Mas pra levar tudo de foguete, gasta uma grana preta!”
ELON MUSK: “É aí que tá o pulo do gato.”
REPÓRTER: “Como assim?”
ELON MUSK: “Precisa reduzir drasticamente o custo de lançamento.”
REPÓRTER: “Ah, sei...”
ELON MUSK: “Se o transporte pro espaço ficar barato, as opções se abrem.”
REPÓRTER: “Ou seja, foguetes e IA tão todos interligados!?”
ELON MUSK: “É nesse futuro que eu tô pensando.”
REPÓRTER: “Cara, essa visão já não é de humano da Terra”
ELON MUSK: “Tem possibilidades fora da Terra também.”
joker.icp retweeted
“If something burns your soul with purpose and desire, it’s your duty to be reduced to ashes by it.
Any other form of existence will be yet another dull book in the library of life.”
- Charles Bukowski
joker.icp retweeted
币圈吹了整整三年的“消灭助记词神话”,在今天彻底沦为了一场血腥的笑话。
几个小时前,链上安全圈拉响了最高级别的红色警报:
为海量 Web3 钱包提供“谷歌/苹果一键登录”的底层基础设施 Web3Auth,突发私钥泄露级严重事故。
首当其冲被穿透的,是 Cosmos 生态龙头钱包 Keplr。
黑客已经开始在受影响的地址上发起疯狂盗币。
Keplr 官方紧急切断所有社交登录入口,全网各大链上验证节点和项目方一片大乱。
极度荒诞的一幕在链上赤裸裸地上演:
那些自以为走在时代前沿、贪图省事用 Google 和 Apple 账号一键建钱包的散户,账户私钥被扒得一干二净,沦为案板上的鱼肉;
反而是那些被嘲笑“操作落后、冥顽不灵”、坚持拿纸笔手抄 12 个助记词、死守冷钱包的老韭菜,在这场灾难里毫发未伤。
为了省下抄写助记词的 30 秒,代价是把自己的全部身家,双手奉送给第三方中心化数据库。
现在的链上战场已经变成了一场生死竞速。
Web3Auth 联合 Keplr 正连夜在 Celestia、Injective、dYdX、Osmosis、Akash 等 10 条公链以及 8 条 EVM 链上,疯狂用白帽程序把被盗边缘的资产转移到紧急恢复钱包,硬生生跟黑客抢跑。
Polygon 首席安全官 Mudit Gupta 更是公开发出最冰冷的警告:
“只要你用社交账号登过 Keplr,现在立刻马上把钱转走!更可怕的是,天知道全网还有多少钱包和 dApp,也在背后接了这家服务商。”
Not Your Keys, Not Your Coins。
你以为一键登录是去中心化的技术进步,其实你只是自作聪明地,把钱存进了一家没有任何安保的地下钱庄。
joker.icp retweeted
There is NO WAY GPT-6.2 Sol just beat Opus 5.5.
After some back-and-forth on, I was finally able to get my hands on the routed GPT-6.2 Sol through a friend.
I tested it against Opus 5.5 using the exact same prompt, both at their highest reasoning settings, and the results genuinely surprised me.
Here's how they compared:
- GPT-6.2 Sol: Took around 20 minutes and cost $9.
- Opus 5.5: Took 1 hour and cost $20.
Sol nailed the details really well and produced something you'd expect to see in a low-budget animated movie.
Opus 5.5 focused more on the overall environment, and it looked really good too. But I have to be honest, Sol did better here.
Which one do you think won this comparison?
Fable 5.5 just f*cked the whole mecha anime industry...
this is f*cking crazy
i made a giant robot catch a meteor with Fable 5.5, no image generation, no video model, 100% code
> 6 agent teams working in parallel
> an AI art director on every team
> 36 shots, 1,968 code-rendered frames
> zero image generation, zero video models
> every frame coded, not drawn
full workflow + code in the comments
then, read the complete motion design guide
joker.icp retweeted
An OpenSSH bug that took 46 years to ripen!
1980
UNIX puts * in the password field of system accounts. The idea is simple: it’s not a real hash, so it can’t be brute-forced and you can’t log in with a password.
2012
The musl libc developers decide that returning NULL from crypt() on failure is dangerous - programs often forget to check it and crash with a segfault. So they return an error string instead.
Guess which one?
The same asterisk: *.
Three weeks later they add a length limit: passwords longer than 256 bytes burn CPU. Anything longer is rejected. And since it’s an error, crypt() returns *.
2016
OpenSSH fixes a timing attack (CVE-2016-6210). If a user doesn’t exist, the server still hashes the password with a fake/borrowed salt so an attacker can’t tell valid usernames by response time.
2022
Alpine Linux changes the locked root password in /etc/shadow from ! to * so SSH keys work properly on cloud images.
… and the loop closes!
You connect over SSH as root (whose shadow entry is *).
You type any password longer than 256 characters (e.g. 300 As).
crypt() hits the length limit → returns * → OpenSSH compares it with the value in the database → that value is also * → you’re in as root.
That chain took 46 years (!!!) to complete.
Every individual change was reasonable. Together they produced a perfect * == * authentication bypass (CVE-2026-123158, nicknamed *sshole). Fixed in OpenSSH 10.6p2 and Alpine packages.
🚨 Critical OpenSSH flaw allows passwordless account access using any password over 256 characters
sshhole.com
A newly disclosed vulnerability dubbed *sshole has been discovered in OpenSSH installations using musl libc, potentially allowing attackers to bypass authentication and gain access to accounts that should reject password logins entirely.
The flaw affects certain configurations of Alpine Linux, OpenWrt, Buildroot, and Yocto where PAM is disabled and password authentication remains enabled. Under specific conditions, attackers can authenticate to locked accounts using arbitrary passwords between 257 and 1,024 characters.
The researcher demonstrated that affected Alpine Linux cloud images could allow attackers to gain root privileges through a passwordless administrative account.
OpenSSH 10.6p2 addresses the issue. Administrators are urged to update immediately or disable SSH password authentication until patches can be applied.

















![[archer] (@archer_uwu) profile picture](https://pbs.twimg.com/profile_images/1851938840916664320/ocWeeHC2_400x400.jpg)
![Image from [archer] (@archer_uwu)](https://pbs.twimg.com/media/HWo1PyyWsAAT6as.png?name=small&format=webp)












