在一个激烈竞争且“狂热”的AI领域中，投资者莎拉正驾驭着一个由AI模型的递归自改进所推动、并即将迎来指数级智能的世界。这种快速的步伐在研究人员中造成了不安全感和打破既有叙事的信念，其中一些人觉得他们的个人努力被巨大的计算需求所掩盖，从而产生了一种无力感。 莎拉的投资理念深深植根于她所称的“伟大的男人和伟大的女人历史理论”，她相信得到适当支持的“高能动性个体”能够改变结果。尽管她本人不渴望成为“伟大的女人”，但她的目标是通过理解、做出正确判断并赋能非凡的人来成为最好的投资者。她的公司 Conviction 采用“技术先行”的方法，专注于“现在可能”且“有价值”的事物，经常发现非显而易见的机会。例如，她将法律的结构视为“语言”，投资了 Harvey，预见到AI将对法律实践产生变革性影响，从琐碎的任务到复杂的并购工作。同样，她提到了由 Tony Zhao 和 Chang Chi 共同创立的机器人公司 Sunday Robotics，他们数据收集的创新方法和对半人形机器人的宏伟愿景，以其速度和以结果为导向的技术逻辑打动了她。 她的投资过程是凭直觉的：她迅速对人和想法形成初步的积极判断，然后投入大量时间识别自己理解中的不足，在该领域扎根，并寻求外部反馈。这种“攀登至信念”的过程涉及详细的备忘录和合伙人讨论，确保决策基于深刻的直觉而非仅仅是背景。 莎拉强调了当前AI环境中的几个关键挑战。计算能力是一个主要问题；基础设施领域的领导者认为，在2030年之前不会有重大的、能产生实质性影响的变化，这受到监管障碍的阻碍，而不仅仅是技术问题。例如，建设数据中心需要公众和政治上的协同，类似于核能领域的挑战。她担心物理供应链以及与软件相比，实体建设的缓慢速度。另一个担忧是一些投资者缺乏基本理解，他们可能会将判断委托给背景，而非基础的商业和技术理论，从而导致潜在危险的大规模押注。 一场重要的辩论围绕着开源AI模型展开。莎拉认为“覆水难收”，具有竞争力的开源模型正在全球范围内涌现。限制它们在美国的使用只会阻碍守法的美国企业，而敌对行为者则不受影响。她主张对这些模型进行严格的安全测试，而非猜测，以理解和控制风险。在开源支持下，广泛获取“廉价到无法计量的智能”，对于将AI能力普及到整个经济体至关重要，因为没有哪个单一的前沿实验室能想象出其多样化的应用。这引出了“计算独立”的概念，她认为这对于国家安全和经济竞争力至关重要，并敦促投资于替代芯片架构、多元化供应链（例如 Paxilica）和能源。 受她的创业型父母启发，莎拉相信无中生有，专注于价值，善待他人，并与非凡个体合作。她拒绝那种将营销置于实质之上的犬儒主义，并倡导一种正和方法，在这种方法中，合作而非仅仅是零和竞争能够蓬勃发展。 展望未来一年，莎拉希望看到“实践中的杰文斯悖论”——AI智能体更有效地处理日常任务，从而解放人类时间。她预见到一场类似于软件工程的变革，届时个人和公司将变得更高效，带来就业增加而非取代，前提是新的工具和相关教育能够普及。她最后表达了对硅谷那些有成就的个人的深深感谢，他们将想法和人置于背景之上，敢于冒险并给予鼓励，莎拉认为这是她旅程的基石。

In a violently competitive and "frenzied" AI landscape, Sarah, an investor, navigates a world poised on the brink of exponential intelligence, fueled by recursive self-improvement of AI models. This rapid pace creates insecurity and narrative-breaking beliefs among researchers, some of whom feel their individual efforts are overshadowed by massive compute needs, leading to a sense of disempowerment. Sarah's investment philosophy is deeply rooted in what she calls "great man and great woman theories of history," believing that high-agency individuals with the right support can change outcomes. While she doesn't aspire to be a "great woman" herself, her goal is to be the best investor by understanding, being right, and empowering extraordinary people. Her firm, Conviction, focuses on what is "now possible" and "valuable" with a "technology-forward" approach, often identifying non-obvious opportunities. For instance, observing the structure of law as "language," she invested in Harvey, anticipating AI's transformative impact on legal practice, from trivial tasks to complex M&A work. Similarly, she points to robotics company Sunday Robotics, co-founded by Tony Zhao and Chang Chi, whose creative approaches to data collection and ambitious vision for semi-humanoid robots impressed her with their speed and outcome-driven technical logic. Her investment process is instinctive: she quickly forms an initial positive judgment on people and ideas, then dedicates significant time to identifying gaps in her understanding, grounding herself in the domain, and seeking external feedback. This "climb to conviction" involves detailed memos and partner discussions, ensuring decisions are based on deep intuition rather than mere pedigree. Sarah highlights several critical challenges in the current AI environment. Compute capacity is a major concern; infrastructure leader suggests no significant needle-moving changes before 2030, hampered by regulatory hurdles, not just technology. Building data centers, for example, requires public and political alignment, akin to the challenges in nuclear power. She worries about the physical supply chain and the slow pace of building things compared to software. Another concern is the lack of fundamental understanding among some investors, who might proxy judgment to pedigree rather than the underlying business and technical theory, leading to potentially dangerous large-scale bets. A significant debate revolves around open-source AI models. Sarah argues that "the cat is out of the bag," with competitive open-source models emerging globally. Restricting their use in the US would only hinder law-abiding American businesses, while adversarial actors remain unaffected. She advocates for rigorous safety testing of these models, rather than speculation, to understand and control risks. Broad access to "intelligence too cheap to meter," supported by open source, is crucial for diffusing AI capabilities across the economy, as no single frontier lab can imagine the diverse applications. This leads to the concept of "compute independence," which she views as essential for national security and economic competitiveness, urging investment in alternative chip architectures, diversified supply chains (like Paxilica), and energy sources. Inspired by her entrepreneurial parents, Sarah believes in creating something out of nothing, focusing on value, treating people well, and working with extraordinary individuals. She rejects cynicism that prioritizes marketing over substance and advocates for a positive-sum approach where collaboration, not just zero-sum competition, thrives. Looking a year ahead, Sarah hopes to see "Jevons paradox in practice" – AI agents handling mundane tasks more effectively, freeing up human time. She anticipates a transformation akin to software engineering, where individuals and companies will become more productive, leading to increased employment rather than displacement, provided access and education to new tooling are available. She concludes by expressing profound gratitude for the accomplished individuals in Silicon Valley who prioritize ideas and people over pedigree, taking risks and offering encouragement, which she credits as fundamental to her journey.