Chris

这就是借来的认知和自己努力赚来的认知的区别。

根据老黄自己说Nvidia核心护城河是cuda的装机量. 具体来讲市场垄断优势的装机量会形成以CUDA为核心的庞大开发者生态与软件栈所形成的强网络效应与高迁移成本。 Lex Fridman(01:15:02) Exactly. So NVIDIA is now the most valuable company in the world. I have to ask, what is the NVIDIA’s biggest moat, as the folks in the tech sector say? The edge you have that protects you from the competition. Jensen Huang(01:15:20) Our single most important property as a company is the install base of our computing platform. Our single most important thing today is the install base of CUDA. Now, the reason why 20 years ago, of course, there was no install base. But what makes… And if somebody came up with a GUDA or TUDA, it wouldn’t make any difference at all. And the reason for that is because it’s never been just about the technology. The technology, of course, was incredible, visionary. But it’s the fact that the company was dedicated to it, stuck with it, expanded its reach. It wasn’t three people that made CUDA successful. It was 43,000 people that made CUDA successful. Jensen Huang(01:16:17) And the several million developers that believed in us that trusted that we were going to continue to make CUDA 1, 2, 3, 13, that they decided to port and dedicate their software on top of it, their mountain of software on top of it. And so the install base is the number one most important advantage. That install base, when you amplify it with the velocity of our execution at the scale that we’re talking about, no company in history had ever built systems of this complexity, period. And then to build it once a year is impossible. And that velocity combined with the install base, in the developer’s mind, you just go now, take the developer’s mind. From the developer’s perspective, if I support CUDA, tomorrow it’ll be 10 times better. I just have to wait six months on average. Jensen Huang(01:17:16) Not only that, if I develop it on CUDA, I reach a few hundred million people, computers. I’m in every cloud, I’m in every computer company, I’m in every single industry, I’m in every single country. So if I create an open source package and I put it on CUDA first, I get these both attributes simultaneously. And not only that, I trust 100% that NVIDIA is going to keep CUDA around and maintain it and improve it and keep optimizing the libraries for as long as they shall live. You could take that to the bank, and that last part, trust. You put all that stuff together, if I were a developer today, I would target CUDA first. I would target CUDA most. And that’s the reason that I think in the final analysis is our first, that’s even our first-core advantage.

@RobinSeun @grok 瑞达利欧过去20年年化收益率大致是多少？ 其中全天候策略的年化收益率是多少？

雷达里奥老师也开始发力了。 全天候投资组合我记得是在他的原则里提过的他看家的本领。 现在也拿出来分享了，存下来慢慢看。 建议这种文章自己看，别ai解读。

遭受硅谷之劫的企业家们从中学到4点经验，这些经验直到今天仍在主导商业思想： ①循序渐进。 不能沉溺在宏大的愿景中，否则会使泡沫膨胀。自称可以成大事的人都不可信，因为心存改变世界之雄心的人通常要更加谦逊。小幅地循序渐进地成长是安全前进的唯一道路。 ②保持精简和灵活性。 所有的公司都必须留出一定空间，不要事事都严格计划。你不知道你的事业会变成什么样，事先规划通常既死板又不现实。相反，你应该做些尝试，反复实践，把创业当成未知的实验。 ③在改进中竞争。 不要贸然创造一个新市场。以现成的客户作为出发点创业才更有保障。成功者已经创造出被认可的产品，在此基础上加以改进，才是可取之道。 ④专注于产品，而非营销。 如果你的产品需要广告或营销人员去推销，就说明你的产品还不够好：科技应用于商业应该主打产品开发，而不是分销。在泡沫年代打广告显然都是浪费，唯一持久的成长是爆发式成长。 这些经验教训在创业领域成了信条；忽视它们的人被认为会遭受2000年美股大崩盘、科技股重挫那样的厄运。然而这些法则的对立面可能更正确： 1.大胆尝试胜过平庸保守。 2.坏计划也好过没有计划。 3.竞争性市场很难赚钱。 4.营销和产品同样重要。 #从0到1

大多数公司能做好这其中一两点就很难了，这可能就是大多数公司没那么成功的其中几个原因。一般公司会逐步扩大，默认就是招更多人，裁人是逼到无奈才开始开除人，随着规模扩大越来越多不懂技术的人参与决策，随着不懂技术的人加入需要对其认知，无效会议自然变多，公司老大甚至不懂技术🥴。

@linexjlin @ivanalog_com 感谢 看起来地月L4和L5确实不错，日照时间接近100%，RTT在两三秒钟，能满足绝大部分推理需求，离地球近一点散热面积可能得搞的比日地L4 L5大一些。

@chris___sun @ivanalog_com 地月，日地拉格朗日点： x.com/yifan_zhang_/s…

The tradeoff here is between chasing what’s exciting and building what’s durable. The founders who are thriving right now stopped caring about model capabilities and started caring about the things models can’t take away: data moats, workflow capture, integration depth. It’s less fun to talk about at a dinner party. It’s where the actual companies get built.

1. 大规模数据中心推理所占用的算力大约在70%~90%，训练占10～30%，根据研究和应用类型不同有些许差距，大头还是推理。 3. 太空中太阳能无大气层和地球昼夜变化(离地球足够高)，数据中心在太空更容易散热，效率将是地面的10倍以上，并且更容易规模化。 如果是地球同步轨道的数据中心到地球表面延迟大约120ms，延迟是可以被感知到的，未来可能太空和地面相互补充，地面负责实时推理需求而太空负责非实时推理需求。

@ivanalog_com 1. 深度推理/训练都对延迟并不敏感 2. 北纬42-50度的散热优势，不见得有能源优势，能源传输还有损耗 3. 还有就是地球的能源总量有限，长期来看太空数据中心可能更合适。

Ai agent 和传统软件开发相比，虽然仍有大量安全性和不确定性的瑕疵，但想象空间最大的地方是： 一， 可以自己改进自己的源代码 二， 可以自发的寻找新的方法来实现某个目的 三， 可以自发的在开放的空间和外界其它 ai agent 进行复杂的价值交换，来完成以前传统软件根本无法完成, 也无法想象的任务。 尤其是第三点，如果在支付手段上达成共识，大量ai agent 自发的，不间断的协作，演化，这就有生命体的雏形了. 目前的 open claw 之类的工具还属于玩具，但他有潜力在现在的硬件 os 之上构建一个抽象层，变成用户和硬件交互的主要方式. 这还了得，微软和苹果绝对不可能允许这种事情发生，肯定会构建自己的 ai agent 生态，依靠自己的强大分发能力把竞争者边缘化。 尽管如此，绝大部分开发者最终结局仍然只是凑热闹的炮灰，因为他们并没有一丁点垄断性的优势，纯粹靠 ai agent 和热钱带来的多巴刺激在维持着。 未来大概率会有现在无法想象的新的商业模式可以成功捕捉价值。就像微电脑革命始于 1975年，但你完全可以冷眼旁观，躺平十一年到 1986年微软上市后再出手，再躺平 40年收获四千倍的回报。

It might be the caffeine speaking, but I love you all ❤️ Happy Valentine's day you sexy mf'ers. I'm writing this at a coffeeshop. Rain outside. A first date happening next to me, I'm hoping they make it. I'm feeling happy & grateful for all of this, human civilization, life on Earth, a spinning rock in space. There are 100-400 billion other planets in our galaxy alone. And there are ~2 trillion galaxies in just the observable universe. And somehow we have a chance to crack open the mysteries of the universe, to solve physics, biology, intelligence, to understand our own mind, to travel out toward the stars. And at the same time, we often bicker about the stupidest shit. The whole thing is hilarious and beautiful. All of this is a miracle ❤️ PS: It's definitely the caffeine speaking 🤣

"Reality is the best verifier." dwarkesh.com/p/elon-musk

With the iteration of AI, it may be better in person this year or next year, but it is unlikely that people will improve much in one or two years. People are too difficult to change, and they must say something that they don't know in order to make themselves appear as if they know it.

"Really what will happen is that the pure AI, pure robotics corporations or collectives will far outperform any corporations that have humans in the loop. And this will happen very quickly."

true.

true 🤣

Ralph is a primitive, I also felt that there is a middle ground where I want to have an agent that can delegate work to Ralph's, I tried GSD, but it was a bit cumbersome and didn’t quite fit well with my current workflow. So, I manually managed my memory to assign tasks to Ralph’s workflow, which was very similar to the Villager project. Next, I plan to try using beads to help manage my memories, which might be more effective.

Once a product is "good enough" to solve your problem, more tech doesn't make it better—it just makes it more complex. If everyone is way over-delivering on performance, their "different advantages" aren't actually different anymore. #TheInnovatorsDilemma

借来的智慧，和自己挣来的智慧，有很大区别。前者在压力面前会崩溃。

incumbents rationally ignore “small, low-profit” markets, and that’s exactly where disruptive technologies get strong enough to later beat them in the mainstream.

I’ve been doing this for some time now: decoupling different tasks and identifying their dependencies. This allows me to work on multiple tasks simultaneously. Most importantly, each task is independent, making it easier for agents to verify them. This approach also reduces the likelihood of hallucinations and results in better outcomes.