在伦尼的播客上一次引人深思的讨论中，A16Z专注于消费投资的普通合伙人安尼什·阿查里亚，对人工智能的未来提出了一个乐观而广阔的看法，挑战了普遍存在的担忧，并强调了人工智能放大人类抱负和能动性的潜力。 阿查里亚直接驳斥了硅谷中普遍存在的关于人工智能将创造“永久底层阶级”的“有趣而黑暗的幻想”。他认为这种担忧大都没有根据，并引用证据表明机会比以往任何时候都更加分散，网络效应不再那么中心化，并且即使是像放射科医生和程序员这样被认为容易受到影响的职位，其招聘信息数量仍然很高。他区分了“递归自我改进（RSI）”和“自催化效应”，指出当前的人工智能进步并未导致失控的赢家，而是促进了许多参与者之间的流程改进。他还质疑有多少问题是真正“受智力限制的”，暗示许多现实世界的挑战受到其他因素的限制。 在公司内部，阿查里亚看到了对人工智能的积极拥抱，从卡瓦克（Kavak）的机械师到谷歌（Google）的高管，员工们都在使用新工具来提高生产力。他指出，人们正在从简单地“使用人工智能”转向“围绕人工智能重组整个公司”，将其比作电力的缓慢普及和最终对工厂的重新设计。最有抱负的公司正在围绕模型重新思考一切。 阿查里亚引入的一个核心概念是，公司建设将越来越成为“一系列创造循环”。他解释说，就像编码代理将模型与工具和内存结合成一个循环一样，业务功能也将采用类似的循环结构。这些循环可能从个人层面延伸到整个公司单元，自动化诸如错误修复、增长实验或营销活动之类的任务。然而，他强调人类仍然是关键要素：“这个循环将帮助你爬到局部最优值，但随后就会停滞。你需要人类的直觉。你需要有人真正帮助你找到下一座山的底部。” 这种愿景暗示了一种代理与人类的反馈循环，其中人工智能负责攀爬，而人类则识别下一座“山”或战略方向。 阿查里亚认为，人工智能将导致模型使用的分化：“前沿模型”（如阿斯特拉）将保留给具有无限上升空间的角色（例如药物研发、销售、研究、复杂工程），而“开放权重”或“中等智商”模型则用于效率和成本效益至关重要的任务（例如日常法律、财务或客户支持任务）。他将自己描述为“模型品鉴师”，强调不同的模型具有独特的“形态”和专业化——有些擅长创造力（例如用于讲故事的奎恩 3-8 Max），而另一些则提供神经质的精确性（例如 GLM 5-3）。 他强烈倡导亲身实践人工智能，敦促所有人“只管动手做”一些东西。他建议选择一个非关键项目，并将其作为一个“底盘”来学习和培养直觉，强调“建造现在就是阅读”。他关于儿子“破解”他的AI驱动的屏幕时间奖励系统的个人轶事，凸显了这些工具所带来的创造性以及有时不可预测的互动。 超越生产力，阿查里亚在消费人工智能中看到了巨大的机遇，其核心理念是“循环，让我更快乐”。他认为人们渴望“花时间”而非“节省时间”，并且人工智能可以满足人类的基本需求，比如感觉更亲密、被爱、取得进步和获得乐趣。他指出，挑战不在于模型或能力，而在于产品设计，敦促创始人探索“我们社会存在中不舒服的部分”。 阿查里亚对人工智能更广泛的社会影响表示乐观，呼应了它“恰逢其时”以应对全球挑战的观点。他认为它可以让“重要的事情变得便宜”，尤其是在医疗保健和教育领域，这些领域已变得极其昂贵。他还强调了人工智能放大个人身份和能动性的潜力，将“技能与愿望解绑”，从而让更多人追求他们的创造性抱负。 关于公司的持久性和护城河，他引用了来自迪卡贡（Decagon）的杰西的话：“护城河通常是被发现的，而非设计出来的。” 他提醒听众，经典的护城河（网络效应、规模优势、品牌、专有数据）仍然有效。他认为当前环境对初创公司更有利，因为创新之门“大开”，对专业化产品有很高的支付意愿，并且能够在现有企业避免的方向上进行建设。 他给产品人员的建议是：“只管动手做。”他建议每周发布一些东西，无论多么小或看似不重要，以培养直觉和掌握能力。他重申，硅谷奖励脆弱性和正和思维，鼓励人们参与并分享他们所构建的东西。 安尼什·阿查里亚的观点描绘了一个未来，人工智能赋能个人，通过级联的“循环”改变行业，并解决人类深层次的愿望，前提是我们带着抱负、好奇心和建设的意愿来对待它。

In a thought-provoking discussion on Lenny's podcast, Anish Acharya, a General Partner at A16Z focusing on consumer investing, offered an optimistic and expansive view on the future of AI, challenging common fears and highlighting its potential to amplify human ambition and agency. Acharya directly addressed the "funny, dark fantasy" prevalent in Silicon Valley about a "permanent underclass" being created by AI. He argued that this fear is largely unfounded, citing evidence that opportunities are more distributed than ever, network effects are less centralized, and job postings remain high, even for roles like radiologists and programmers thought to be vulnerable. He distinguished between "recursive self-improvement" (RSI) and "autocatalytic effects," suggesting that current AI advancements are not leading to runaway winners but rather to improved processes across many players. He also questioned how many problems are truly "intelligence-bound," suggesting that many real-world challenges are limited by other factors. Inside companies, Acharya sees a positive embrace of AI, with employees, from mechanics at Kavak to Google executives, using new tools to enhance productivity. He noted a shift from simply "using AI" to "reorganizing entire companies around AI," comparing it to the slow diffusion of electricity and the eventual redesign of factories. The most ambitious companies are rethinking everything around models. A core concept Acharya introduced is that company building will increasingly become a "series of creating loops." He explained that just as coding agents involve models in a loop with tools and memory, business functions will adopt similar loop structures. These loops could range from personal to entire company units, automating tasks like bug fixes, growth experiments, or marketing campaigns. However, he stressed that humans remain a critical ingredient: "The loop will help you climb to the local maxima, but then it plateaus. You need human intuition. You need somebody to actually help you land at the base of the next hill." This vision suggests an agent-to-human feedback loop where AI handles the climbing and humans identify the next "hill" or strategic direction. Acharya believes that AI will lead to a split in model usage, with "frontier models" (like Astra) being reserved for roles with unbounded upside (e.g., drug discovery, sales, research, complex engineering), and "open weight" or "mid IQ" models being used for tasks where efficiency and cost-effectiveness are paramount (e.g., routine legal, finance, or customer support tasks). He described himself as a "model sommelier," emphasizing that different models have unique "shapes" and specializations – some excel at creativity (like Quen 3-8 Max for storytelling), while others offer neurotic precision (like GLM 5-3). He strongly advocates for hands-on experimentation with AI, urging everyone to "just make" things. He suggested picking a non-critical project and using it as a "chassis" to learn and build intuition, emphasizing that "building is now the reading." His personal anecdote about his son "hacking" his AI-powered screen time reward system underscored the creative and sometimes unpredictable interaction with these tools. Looking beyond productivity, Acharya sees a massive opportunity in consumer AI centered around the idea of "loop, make me happier." He posited that people desire to "spend time" more than "save time," and that AI can address fundamental human needs like feeling more connected, loved, making progress, and having fun. He argued that the challenge isn't in models or capabilities but in product design, urging founders to explore "uncomfortable parts of our social existence." Acharya expressed optimism about AI's broader societal impact, echoing the sentiment that it arrived "just in time" to address global challenges. He believes it can make "important things cheap," especially in healthcare and education, which have become prohibitively expensive. He also emphasized AI's potential to amplify individual identity and agency, "unbundling skill from desire" and allowing more people to pursue their creative ambitions. Regarding company durability and moats, he cited Jesse from Decagon, stating, "moats are most often discovered, not designed." He reminded listeners that classic moats (network effects, scale advantages, brand, proprietary data) are still valid. He believes the current environment is easier for startups due to open "floodgates" for innovation, high willingness to pay for specialized products, and the ability to build in directions incumbents avoid. His advice for product people: "Just make." He suggested shipping something once a week, no matter how small or seemingly unimportant, to build intuition and mastery. He reiterated that Silicon Valley rewards vulnerability and positive-sum thinking, encouraging people to engage and share what they've built. Anish Acharya's perspective paints a future where AI empowers individuals, transforms industries through cascading "loops," and addresses deep human desires, provided we approach it with ambition, curiosity, and a willingness to build.