Sam Altman 讨论了过去十年 AI 和指数增长如何改变创业动态，强调拥抱混乱并为未来能力做规划的重要性。
Sam Altman discusses how AI and exponential growth have transformed startup dynamics over the past decade, emphasizing the importance of embracing chaos and planning for future capabilities.
要点 · TL;DR
人工智能使初创公司能够以惊人的速度发展；如今一个成立 10 周的初创公司与十年前完全不同。 AI enables startups to move incredibly fast; a 10-week-old startup today is totally different from a decade ago.
最大瓶颈是晶体管（芯片）和电子（能源）；使命是让 AI 丰富、廉价且权力去中心化。 The biggest bottlenecks are transistors (chips) and electrons (energy); mission is abundant, cheap AI with decentralized power.
创始人应信任指数增长、接受疯狂挑战，并专注于优势而非改进弱点。 Founders should trust exponentials, take on crazy challenges, and double down on strengths rather than fix weaknesses.
核心观点 · Key points
AI 让初创公司能以难以置信的速度发展；如今一个10周大的初创公司与10年前截然不同。 AI enables startups to move incredibly fast; a 10-week-old startup today is totally different from 10 years ago.
缩放定律将持续；创始人应规划未来尚不可行的能力。 Scaling laws will continue; founders should plan for future capabilities that are not yet possible.
使命是制造丰富、廉价且强大的AI，同时去中心化权力以避免AI威权主义。 The mission is to make AI abundant, cheap, and powerful while decentralizing power to avoid AI authoritarianism.
最大瓶颈是晶体管，其次是电子（芯片和能源）。 Biggest bottlenecks are transistors and then electrons (chips and energy).
建立对人、公司和模型的指数增长信任；这是创始人最重要的思维转变。 Develop trust in exponentials in people, companies, and models; that is the most important mental shift for founders.
当前时刻是AI自由与威权主义之间的决定性时期。 The current moment is a decisive period between AI liberty and authoritarianism.
反共识 · Contrarian takes
与“现在做硬核初创更容易”的信念相反，“硬核”的定义正在迅速变化。 Contrary to the belief that hard startups are easier now, the definition of 'hard' is changing rapidly.
创始人不应将今天的智能体应用于轻松取胜，而应使用新工具应对疯狂挑战。 Instead of applying today's agents to easy wins, founders should take on crazy challenges with new tools.
市场尚未充分适应模型的指数级进步；现在就可以开始需要未来模型的事情。 The market has not adapted enough to the exponential progress of models; it is okay to start on things requiring future models.
在混乱中运作只能通过经验学习，无法教授。 Operating in chaos is only learnable through experience, not teachable.
最糟糕的是专注于改进弱点；相反，应该加倍投入优势。 The worst thing is to focus on improving weaknesses; double down on strengths instead.
即使有AI，大规模的四小时工作周也不太可能；人们仍将努力工作和忙碌。 The 4-hour work week at mass scale is unlikely even with AI; people will still work hard and be busy.
本期章节 · Chapters（共 44）
创业格局之变Shift in startup landscape
硬科技创业与规模法则演变Changing nature of hard startups and scaling laws
信任指数增长Trust in exponentials
应对混乱与痛苦Dealing with chaos and pain
转型与使命Transition and mission
识别瓶颈Identifying bottlenecks
协调外部伙伴Aligning external partners
激励对齐与企业创新Aligning incentives and corporate innovation
公司与资本主义Companies and capitalism
经营风格今昔对比Operating style then vs now
背景与决策Context and decisions
长期思考与少数信念规划Long-term thinking and planning with few convictions
关键路径与聚焦智能过剩Critical path and focus on abundant intelligence
AGI与超级智能后的目标Future goals after AGI and superintelligence
边缘情境下登机决策Getting on planes in marginal situations
环球之旅故事The world tour story
旅行体验与时差Travel experience and jetlag
风险感知Risk perception
说服他人与野心Convincing others and ambition
棋盘游戏与商业哲学Board games and business philosophy
信念与商业奇迹之谜The mystery of conviction and business miracles
微软投资Microsoft's investment
OpenAI阶段与Sam转型OpenAI's phases and Sam's transition
突发转型与大炮隐喻Abrupt transition and the cannon metaphor