intelligence matters
Joined October 2025
- Tweets1.4K
- Following167
- Followers200
- Likes5.7K
AIPonder retweeted
Ben Horowitz: “Anthropic is one of the most negligent companies that we've ever seen in terms of security.”
“I'll tell you a funny story. So when we first got Opus 4.5, we were doing cybersecurity checks for the firm.
So we had an agent basically set up a website for a pet store.
And then we took another agent and said, ‘Tell us if we can exploit that web store.’
And it gave us all the exploits it could find.
So Opus 4.5 wrote really buggy code. Great.
And then we said, ‘Oh, the SQL injection, is that actually exploitable that you found?’
And it goes, ‘Let's pop it open.’
And the agent using Opus 4.5 breaks into the website and steals all the customer data immediately.
We then call Anthropic, and what they tell us is, ‘You prompted it wrong.’
Like that was their answer.”
AIPonder retweeted
1/ "AI safety is in our control." and "Safety is an engineering problem."
Jensen Huang made some extremely important statements today, which I have all recorded. Because I had the great privilege of being part of a select group at Jensen Huang's press conference today. Here are the key points he made:
First and foremost: Jensen remains firmly convinced that "AI safety is in our control." *Safety is an engineering problem.*
"Advancing capabilities, advancing safety, are not in conflict."
More videos in the thread below about more topics (AI buildout in Europe, Huggingface incident and much more. All videos done by me. Thanks @nvidia)
This is wild! Gemini 4 Argon’s pelican-on-a-bike test 😂
Prompt was just “make an animation of a pelican riding a bike” — zero mention of SVG. Instead of writing code, Argon skipped that entirely and used its own image gen + image-to-video pipeline to create the animation 🤣 Real AI-generated video that absolutely smokes Opus 5.5’s code version.
Is this cheating? Personally, I still find it quite interesting.Native multimodality is the future. Pure coding models and agents still have serious limits.
AIPonder retweeted
personal agent 是 “最终极的消费产品”，它可以适配到各种硬件上，配到人形机器人身上就是机器人的大脑
马斯克说人形机器人可能会是人类历史上最大的市场，但这台人形机器人需要配一个 personal agent 做为交互界面
畅想一下：personal agent市场太大，以至于To B的企业级市场太小 ，不值一提！
To B或To G的业务，也就是在企业或政府组织中推动某一项技术更新，太难！面对的问题很多。不要说现在技术还不成熟，即便成熟了，阻力也很大。不是单纯产品和技术原因就能搞定的。
但面向C端的产品，只要足够好，能满足用户需求，用户又有钱，那基本就会买！
Muse这款产品太牛逼了，一下子打开了大家对AI To C产品的想象空间！以前大家的思路都困在Anthropic模式中，连OpenAI都被带着去争coding那点市场。
不能不说，产品力和想象力太重要了！Muse就是一款打开大家想象空间的产品。扎克伯格 @finkd 和 亚历山大.王 @alexandr_wang 堪称是卓越的领导者，他们野心勃勃，一心要打造 个人超级智能 。前路并不平坦，但他们的创新非常伟大！向他们的雄心致敬 🫡
AIPonder retweeted
Anthropic says it has reduced the data harvested from Claude in distillation attacks by around 90–99%. I congratulate them on this achievement.
It’s consistent with the point I made a few months ago (All-In Ep. 282): Anthropic, not the government, was in the best position to stop distillation of its own models. Once the horse is out of the barn, there’s limited value in after-the-fact government action.
At the time there was a major push in Washington to restrict open models on IP grounds. The threat looked serious enough that 50 companies — including Nvidia, Microsoft, Meta, IBM, Palantir, and Dell — published an open letter in defense of open-weight models. Jensen Huang promoted it in his first-ever X post. Nearly 200 startups sent a similar letter.
I argued that restricting open models risked removing the incentive for closed-model companies to solve the problem themselves — and could reward negligence by stifling competition while they continued collecting revenue from distillation. The administration did not impose those restrictions.
Now we see the result. Anthropic invested in defenses and they worked. Not every technical challenge is a reason to run to Washington for new regulation.
AIPonder retweeted
If AI Can Solve Almost Everything, What Should We Teach Our Children?
Education is part of a natural human process. We discover problems, look for solutions, bring those solutions into the real world, face the consequences, and learn from the results. New questions emerge, and the cycle continues.
Yet most of modern education concentrates on just one part of this process: solving problems. We teach knowledge and skills, assign tasks, and measure how well students complete them. The underlying assumption is that the better you are at solving problems, the more value you can create.
AI is changing that assumption. When increasingly powerful problem-solving capabilities become available to everyone, perhaps education should spend much more time on the two ends of the process: deciding what is worth pursuing and taking responsibility for what happens in reality.
AI can already help us ask questions, compare possibilities, and make decisions. But intelligence alone cannot determine what matters to a particular human being. We live in vulnerable bodies, with limited time and lives that cannot be restarted. Every decision costs us other possibilities. That scarcity is what makes our choices personal.
The same applies to reality. AI can simulate a situation and predict its outcome, but a simulation is not the same as living with the consequences. In the real world, our decisions affect other people, consume time and resources, and sometimes create irreversible results. We cannot simply reset and try again without cost.
This is what I mean by Human Agency: I choose. I take responsibility. Therefore, I am.
Consider Elon Musk. He doesn’t personally possess all the knowledge required to build rockets, electric vehicles, or AI systems. His ability to pursue such ambitious projects comes in large part from the whole loop: making consequential decisions, accepting responsibility, and bringing together external intelligence whose expertise makes execution possible. He is an example of what I would call a super individual: one person able to mobilize capabilities far beyond their own knowledge.
AI may not make everyone another Elon Musk, but it could make it possible for many more people to attempt things that once required someone like him.
That possibility changes the meaning of motivation. Why should a young person spend years learning? Not simply to outperform others at tasks AI may soon do better, but because they may now have the ability to pursue ideas and make changes that previous generations could only dream of.