huihsu

欢迎大家共同开发

YOUR ENGINEERING EVOLUTION: Age 20 – Curiosity Age 21 – Trial and Error Age 22 – Overengineering Age 23 – Debugging Age 24 – Fundamentals Age 25 – System Thinking Age 26 – Optimization Age 27 – Design Principles Age 28 – Abstraction Age 29 – Architecture Age 30 – Integration Age 31 – Reliability Age 32 – Scalability Age 33 – Efficiency Age 34 – Elegance Age 35 – Automation Age 36 – Innovation Age 37 – Standardization Age 38 – Mentorship Age 39 – Mastery Age 40 – Legacy

@imwsl90 128G m5 max + dgx spark + 5090

不是决策瘫痪，而是他么穷啊 最近这两天一直在思考是买 64G的 mac 14 pro，还是 48G 的 mac 16 pro 想来想去，发现无法做决策 后来觉得自己是不是出现了决策瘫痪的问题 最后，才明白了 这他妈的不是什么决策瘫痪，这他么是穷啊 随即关掉了页面....

Xchat来了的话，下一步难道是Xvideo

AI时代，Capex才是护城河了

一群细胞，没有实验室、先进制造业、 物流车队、矿业却能够发现、并利用新的元素，合成新的化合物

My brain broke after reading this paper 🤯! Quantum computing just took a serious step toward breaking classical AI limits. MIT Scientists just showed that small quantum computers can outperform EXPONENTIALLY larger classical systems on massive datasets. In a new study, researchers introduced a framework called Quantum Oracle Sketching, where instead of loading entire datasets, the system streams random classical samples, processes them in quantum superposition, and then discards them. This directly bypasses the biggest bottleneck in quantum machine learning, which is data loading 👀 The results are honestly kind of ridiculous. A system with fewer than 60 logical qubits can perform core machine learning tasks that would require exponentially larger classical memory, translating to roughly 10,000× to 1,000,000× smaller system size. They validated this on real data, including single-cell RNA sequencing and movie review sentiment analysis, and showed it works across fundamental tasks like classification, dimension reduction, and linear algebra. They also used Classical Shadows to efficiently extract useful outputs from quantum states, solving another long standing limitation. This matters because modern AI is basically brute force, more data, more GPUs, more cost.

人类要这么建这么多数据中心、这么多加速计算做什么？生成点跳动的矩阵光线欺骗下自己的视网膜接着分泌点多巴胺欺骗下大脑吗？

生成软件依然是保守的AGI，因为前提是假设未来的计算架构还是由目前的软件组成的而不是矩阵。

SpaceX的优势在于，当人类还在用“地表最强”来形容SOTA时直接跳出这个体系

One of the best resources on low-level optimizations agner.org/optimize/micro…

broadcom.cn/blog/myths-of-… 看来需要私有协议、定制化的技术、依然是个巨大的谎言，在GPU互联领域，互联网连接就够了，力大砖飞

AI时代比token freedom更重要的是cognitive freedom

The Fastest Analog Computer Ever Built is A Piece of Glass. Does a lens compute a Fourier transform at the speed of light? A lens doesn’t just focus light. It reshapes the phase of whatever wave you send into it, and as that wave propagates, its structure is reorganized into a new pattern at the detector. Change the input beam, and the output changes instantly.

所以说在这么多人vibe的前提下、蒸馏开源代码的前提下，Anthropic终于炼成万魂幡。 于是用户就变成了消耗token算力可有可无的蛀虫

随便谁掐到一个海峡就可以收费是吧，马六甲呢

token还是每秒个位数的世界 和未来token每秒成千上万上亿的世界截然不同 不可能一直是56k拨号上网时代

“2016 年， AGI的提出者、DeepMind 创始人哈萨比斯曾组建量化交易团队，试图给当时想从 Google 独立的 DeepMind 创收，结果没赚到钱。 “同一年，浙大本硕毕业的梁文锋做量化投资已经 8 年。”

感觉未来员工离职，都不需要交接工作，当天就把电脑收回来，装个小龙虾，根据代码库和电脑上资料生成skill，钉钉飞书账号直接加一个（已离职） 可以继续写代码，飞书也会回复 赛博上班

永远不用技术恐慌的法门 如果这件事情可以轻而易举的解决，那么就不用担心 如果这件事情解决不了，那么也不用担心 如果这件事情可以解决，但是极为困难，说明其中有无数的机会，创造巨大的工作机会，那么也不用担心