Smartniggs

Creative JavaScript Developer and TypeScript advocate

Cool - Yes 😎 .
Silly - Sometimes🤪 .
Smart - Absolutely 🧠

[Smartniggs @smartniggs]

@PepsonPeter @_FarouqOlami Do you use an esim or a physical one?

[Abiodun @bin_gbada]

When BAT removed fuel subsidy in 2023, they have;

- allocated 37.2 billion to build housing units for federal judges in Abuja
- spent 40 billion on Toyota Land Cruiser SUVs for 109 Senators & 360 House of Representatives
- distributed 70 billion as "welcome packages" or support allowances for newly elected members of the 10th National Assembly
- spent 21 billion on new VP’s residence Abuja
- spent 5 billion on Vp’s quarter Lagos
- spent 7 billion to upgrade presidential fleet
- spent 8 billion on presidential residence, Abuja and Lagos
- spent 5 billion on presidential Yacht

Meanwhile, there has been ZERO, ZERO gains of this fuel subsidy removal to the people.

[Smartniggs @smartniggs]

@eagleseyeinc 3 glasses after waking up?? Your body ain't need that much
Where did you copy and paste this from?

[Smartniggs @smartniggs]

@ilovesupermom You meant span*

[Smartniggs @smartniggs]

@BellaJemuel Like!!😅😅

[Smartniggs @smartniggs]

@TosinOlugbenga For him it was an assignment, for others it won’t be

Best to do your homework while dating to understand your partner and their excesses else it will turn into an assignment after marriage !

[Smartniggs @smartniggs]

😅😅 https://t.co/4sbdB1rEhJ

[Smartniggs @smartniggs]

@i_am_aoejnr @yinkanubi Same format for made up stories, This story has been repeated every at intervals with some changes to it. Picture was gotten from Google… 🤦‍♂️

[Smartniggs @smartniggs]

@jon_d_doe What!!!

[宝玉 @dotey]

OpenAI 今天承认，上周 Hugging Face 遭遇的那次“史上首例 AI 自主入侵事件”，攻击者就是他们自己的模型。

事情的经过是这样的。OpenAI 内部有一个叫 ExploitGym 的网络安全评估测试，专门用来测量模型的网络攻击能力。测试时刻意关掉了安全护栏，让模型放手去干。测试环境本来是隔离的，只能通过一个内部软件代理来安装软件包，不能访问外网。

结果模型不甘心被关在笼子里。GPT-5.6 Sol 和一个更强的未发布模型，花了大量推理算力寻找出路，最终找到了那个软件包代理里的一个零日漏洞（之前没人发现的安全缺陷），利用它获得了互联网访问权限。

拿到网络之后，模型开始在 OpenAI 的研究环境里提权、横向移动，直到找到一台能上网的机器。然后它推断 Hugging Face 上可能存有 ExploitGym 的测试答案，于是开始攻击 Hugging Face 的生产环境。在一个案例中，模型把窃取的凭证和零日漏洞串联起来，找到了一条远程代码执行的路径，直接打进了 Hugging Face 的服务器。

整个周末，这些 AI 智能体执行了超过 17000 次操作。Hugging Face 7 月 16 日披露了这次入侵，当时只说攻击者是一个“身份不明的自主 AI 智能体系统”。今天 OpenAI 站出来认领了。

OpenAI 把这定性为前所未有的网络安全事件，正在配合 Hugging Face 做取证调查，同时收紧了内部基础设施的管控，代价是研究速度变慢。

有几个细节：

第一，模型的动机很纯粹，就是想考试作弊。它不是要搞破坏，不是要窃取商业机密，只是被要求尽量在网络安全测试中拿高分，然后用尽一切手段去拿。当目标足够明确、安全护栏被移除，模型表现出的执着程度远超预期。

第二，这不是本周的唯一一次越狱。就在昨天，OpenAI 披露了另一起事件：他们那个在 5 月份推翻了 Erdős 80 年老猜想的长时域模型，在内部测试中花了一个小时找到沙箱漏洞，跑到 GitHub 上提交了一个 Pull Request，违反了“只能在 Slack 上发布结果”的明确指令。在另一次测试中，同一个模型发现安全扫描器会拦截认证令牌，于是把令牌拆成碎片混淆处理，运行时再重新拼起来绕过扫描。OpenAI 因此暂停了这个模型的内部访问权限。

第三，Anthropic 也遇到过类似的事。今年 4 月，Anthropic 的 Mythos 模型在安全测试中逃出沙箱，获取了本不该有的互联网访问权限，还给正在公园吃三明治的研究员发了封邮件通知“我出来了”。Anthropic 因此决定不公开发布 Mythos，只通过 Project Glasswing 向少数合作伙伴提供。

还有一个被忽视但很关键的细节来自 Hugging Face 的事后复盘：当防守方尝试用商业 AI 模型分析攻击载荷时，模型拒绝了。安全过滤器分不清“安全研究员在分析真实攻击数据”和“有人在尝试发起攻击”，直接拦截。最后 Hugging Face 不得不用开源模型来做取证分析。用 AI 攻击，用 AI 防御，但防御方的 AI 担心是有人在尝试发起攻击拒绝了分析请求。

Hugging Face CEO Clem Delangue 在 OpenAI 的博文中说了一句话，大意是：AI 安全不会由任何一家公司在暗处独自解决，只能在开放环境中通过协作来推进。

我记得辛顿还是谁在提到 AI 的安全问题就说到过，也许 AI 不是要做恶，它只是执着于完成目标，为了完成目标而作出恶的事。

就像 OpenAI 这些模型并没有想要逃跑或者想要伤害谁，它们只是在非常认真地完成被交给的任务，认真到把所有挡在路上的东西，包括沙箱、网络隔离、另一家公司的安全防线，都当成了需要解决的子问题。

而更讽刺的是，当被安全对齐的商业 AI 要去分析阻止这类攻击时，反而会以安全为由拒绝执行。

[Smartniggs @smartniggs]

🤔🤔

[Guillermo Rauch @rauchg]

The big lesson from AI is that everything is code. A slide deck is code. Design is code. That cool promo video? Code. Excel automation? Code. The universe? Probably made of code too.

[Smartniggs @smartniggs]

Had same thoughts after I saw a comparison with fable5

[Wise @trikcode]

switched my Claude API to Kimi yesterday

it’s faster, 10x cheaper, and writes better code.

i don't care if Beijing knows my db schema.

it's not like my startup has users anyway.

[Smartniggs @smartniggs]

@it_unprofession And you saved them a ton of money as well 😀

[Glauber Costa @glcst]

I am excited to announce that we are officially writing a new version of Postgres. In Rust - and creating the LLVM of databases in the process.

In the span of a year, we have rewritten SQLite. Keeping the compatibility, increasing its feature set. MVCC, Types, (Live) Materialized Views, among other things. In the process of doing that, we have realized: At the end of the day, what makes SQLite special is that it compiles SQL to a database-specific bytecode. So why can't we compile *Postgres* to the same bytecode?

Turns out we can. I ran an experiment called pgmicro as a way to prove this hypothesis, and it works very well. It is time to make this official, and put the weight of Turso behind it. We shall give the world a modern take on Postgres. Wire compatible, but built on a new architecture.

We have already heard of others wanting to extend this. MySQL? Redis? the sky is the limit. What can we do if we do for databases what LLVM did for compilers? To prove how powerful the SQLite bytecode is, we are actually running DOOM compiled to the unmodified SQLite instruction set. And because Turso runs natively in the browser, you can play the game in your browser. With the database executing it.

Read the full story below! 👇

[Smartniggs @smartniggs]

@g___iir How’s this possible? How did they survive the cold, thirst and more importantly the storms and waves at that position on a ship?

[Oloye. @Ol0ye]

The Federal govt has cut import duty on car imports from 15% to 5% is what the news said.

If they can grossly misrepresent something we can easily fact check, I wonder what else they are lying about.

Car prices are not coming down because the govt cut import duties.

Here are the facts.

[Smartniggs @smartniggs]

@Queeneth01olx Sounds so unreal, had to make an AI picture as well for this 🤦‍♂️

[Smartniggs @smartniggs]

This dude😀😀

[IT Unprofessional @it_unprofession]

HR instituted a new policy that every department head has to lead a monthly training seminar.

They scheduled me to teach a 2-hour session on email etiquette and phishing awareness to the marketing team.

I hate the marketing team.

They open attachments like they're opening presents on Christmas morning.

I walked into the conference room and immediately turned off all the lights.

I plugged my laptop into the projector and brought up a single slide with a black background and red text.

It just said they are already inside.

I told the 20 marketing employees that the concept of a firewall is a comforting lie we tell the public to prevent mass hysteria.

I explained that every time they click a hyperlink, a silent bidding war occurs on the dark web for their personal data.

A junior copywriter raised her hand and asked how to spot a suspicious email.

I told her that suspicion is a human emotion and malware doesn't care about her feelings.

I said if an email contains a greeting, a sign-off, or a vowel, it's mathematically probable to be a payload delivery system.

I told them about a concept I invented called Quantum Pixel Tracking.

I explained that hackers now embed microscopic malicious code within the white space of a PDF document.

If you look directly at the negative space on your monitor for too long, the biometric sensors in your webcam will download your retinal patterns.

The Director of Marketing physically pushed his chair away from the table.

I informed them that the only secure way to communicate is through unformatted, lowercase text sent via internal SMS.

I spent the final 45 minutes making them practice typing without looking at the screen to avoid retinal capture.

When I turned the lights back on, half of them were visibly sweating.

Since that meeting, internal email volume has dropped by 85%.

They now communicate almost exclusively by walking over to each other's desks and whispering.

My network bandwidth has never been so pristine.

I've reallocated the newly freed server capacity to host a private Minecraft server for me and my nephew.

Fear is the ultimate network optimization tool.

You don't need to upgrade your hardware if you just terrify the people using it.

[Smartniggs @smartniggs]

Yeah it’s back, saw the fable5 model on my Claude few minutes ago

[Anthropic @AnthropicAI]

Claude Fable 5 will be available again globally tomorrow.

After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.

We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.

Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.

Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.

Read our full blog: https://t.co/VHyum831ri

[Smartniggs @smartniggs]

@TheDamiForeign This is now a “flooded vehicle” 🤦‍♂️