We evolve thoughts so that thoughts could die instead of us

罢工！

i’ve realized that keeping up 100% with new AI tools/models has become almost impossible. the pace is too fast and the breakthroughs hide in tiny details that never get announced. one prompt works today and fails tomorrow. the only people who actually understand where things

开始相信重复可能确实是长期主义的关键。 指望“说一句话”“做一个决策” 就长期解决大部分问题本质上是侥幸懒惰和上岸心理。 如唐僧念经般一直重复一个观点，长期以来对周围人潜移默化的影响是被低估的。

突然想明白股市是不是AI泡沫本质好像并不取决于事实层面是否存在泡沫，而取决于筹码掌控者的共识判断是否有泡沫。在这个基础上去思考是否有泡沫是不是显得有点没啥卵用

What l learned most from studying Duolingo's company culture

Reduce your inflammation, decalcify your pineal gland, read, think and write, think long term, think about bringing value to others, don't be scared to ask, win or fail, learn something new every day, work out, think for yourself, defend your convictions, change your mind.

alex karp said this in $PLTR Q3 Earnings Report: This ascent has confounded most financial analysts and the chattering class, whose frames of reference did not quite anticipate a company of this size and scale growing at such a ferocious and unrelenting rate. Some of our

Elon’s famous analogy: “Every person in your company is a vector. Your progress is determined by the sum of all vectors”. At big companies, vectors often point to different directions. The sum is zero. xAI has a small team of engineers, but every vector points at the same

We’re redesigning @myshell_ai’s organization and product pipeline around AI agents. I want to share few lessons learned to date—and a look ahead. https://t.co/h7eAfCxDGN

曾经的我：OKR 对快速改变方向的创业公司没那么适合！3个月更新一次还是很难与时俱进！ 拆分到每个人头上很容易责权利不清晰引起扯皮！ 现在的我（在意识到企业内 agent 化有多大潜力之后）： OKR管理AI 真好用嘿嘿嘿 总结：执行者会影响你对管理工具有效性的判断

这什么全明星阵容

Much of the mind runs on “shoulds” learned from norms and mimetic desire. We copy what looks desirable, then rebrand it as duty—or worse, rationality. When those “shoulds” overrule what we truly want, the self splits: we live a borrowed life and suffer in the gap.

作为负责人你核心需要的是 1. 确保大的方向和策略是合理的 2. 构建完备的系统和workflow 3. 加速这个系统和现实世界迭代的流程 其他的事情都没那么重要了

你忽略的现实精度足够多， 你在抽象世界理论层面就越无懈可击

your folder can just be your agent bootloader!

@karpathy Long-term, >99% of input and output for AI models will be photons. Nothing else scales. https://t.co/YMOpxp2xtI

“共识”形成的速度之快让我感到费解 大模型的发展不是一道应试题 眼下LLMs有三个大的问题 1，智能和性能是两件事 2，今天模型的“推理”并不是真推理 3，以文本作为唯一物料的世界建模，不够用 从而引申出第四个问题 今年不会出现Agents的爆发，今年依旧是“推理年” 无论哪个模态

https://t.co/mw9vAVpEuG

愈发感觉小公司AI化的快很可能都是被逼出来的。资源永远紧缺，竞争永远激烈，时间线永远紧迫；这些都导致你只有两个选择： 要么激进AI化，要么去死。 在现阶段AI还仍然需要大量文化基建（数据，知识，工作流，云基础设施，懂和敢推进AI的人）的情况下，大公司确实没有超强紧迫性。