Source Video: 原始视频: The Future of Software Development - Vibe Coding, Prompt Engineering & AI Assistants
The central thesis is that infrastructure is constantly layering, and AI (specifically Models) is emerging as the fourth layer. This isn't just an addition; it's a disruptive force. AI is fundamentally reshaping the needs for compute, networking, and storage, and more profoundly, it is reinventing the programming paradigms and application logic of software itself. Software is, for the first time, being disrupted by software. 核心论点是基础设施在不断分层，而AI（特别是模型）正在成为第四层基础设施。这不仅仅是简单的叠加，而是一种颠覆性力量。AI正在从根本上重塑对计算、网络和存储的需求，更深远的是，它正在重塑软件本身的编程范式和应用逻辑。软件，正第一次被软件所颠覆。
This change is a supercycle, similar to the internet or the microchip, creating massive TAM expansion, new user behaviors, and vast opportunities for startups. 这种变革是一个超级周期，类似于互联网或微芯片，带来了巨大的潜在市场规模（TAM）扩张、新的用户行为，并为初创公司创造了巨大的发展空间。
The A16Z Definition: The Technical BuyerA16Z的定义: 技术买家
"Stuff you use to build other stuff." Used by developers, data scientists, DevOps, etc.“你用来构建其他东西的东西。” 被开发者、数据科学家、DevOps等使用。
Vertical SaaS: The Business Buyer垂直SaaS: 业务买家
Used by business professionals in specific industries, e.g., flooring companies or marketers.被特定行业的业务人员使用, 例如地板公司或营销人员。
"We are abdicating LOGIC from the application layer." “我们正在从应用层放弃逻辑。”
Previously, we offloaded resources (compute, storage), but the "what to do" logic always came from a programmer. Now, we ask the model for the answer directly. This forces a rethink of what programmers and software even mean. 过去我们放弃的是资源(计算、存储)，但“做什么”的逻辑总是来自程序员。现在，我们直接向模型索要答案。这迫使我们重新思考程序员和软件的意义。
"Software is being disrupted... it's kind of eating itself." “软件正在被颠覆……它在某种程度上正在自我消耗。”
Go-to-market motions for technical products are radically different from sales-led enterprise software.技术产品的市场进入策略与销售导向的企业软件截然不同。
The firm eventually split its practice into a dedicated Apps Fund and a dedicated Infra Fund to reflect this difference.公司最终将其业务拆分为专门的应用基金和基础设施基金以反映这种差异。
The target audience for developer tools has exploded. With natural language becoming a programming language, this trend is set to accelerate further.开发者工具的目标受众呈爆炸式增长。随着自然语言成为一种编程语言，这一趋势将进一步加速。
On-prem software and perpetual licenses were the norm.本地软件和永久许可是主流。
Shift to recurring revenue. Metrics like Net Dollar Retention became critical.转向经常性收入。净收入留存率等指标变得至关重要。
Accelerated the bottom-up, Product-Led Growth (PLG) trend for developer tools.加速了开发者工具自下而上、产品主导增长 (PLG) 的趋势。
The "most dramatic" shift in the last 30 years, creating entirely new paradigms.过去30年中最具“戏剧性”的转变，创造了全新的范式。
e.g., Compute, Distribution例如：计算、分发
The total addressable market grows dramatically.总潜在市场急剧增长。
Users adopt unprecedented use cases.用户采用前所未有的用例。
Incumbents struggle, creating white space for startups.现有公司难以适应，为初创公司创造了空白空间。
The AI wave is following this exact pattern, creating a massive opportunity. AI浪潮正遵循这一模式，创造了巨大的机会。
Defensibility is evolving from pure technical expertise to include new competitive vectors.可防御性正从纯粹的技术专长演变为包含新的竞争向量。
The lines are blurring. Companies are hard to categorize as just "infra" or "app".界限正在变得模糊。公司很难被简单归类为“基础设施”或“应用”。
In this phase, the market is growing rapidly. "Zero-sum thinking is fatal." There are enough "planets and stars" for everyone to capture value.在此阶段，市场正在迅速增长。“零和思维是致命的。” 有足够的“星球和恒星”供所有人占据。
Eventually, the industry will slow and consolidate into oligopolies or monopolies, which will maintain high profit margins.最终，行业将放缓并整合为寡头垄断或垄断，并维持高利润率。
From demo to real value?从演示到实际价值?
A real boost or a dead end?真正提升还是死胡同?
How far can one model go?一个模型能走多远?
Companionship vs. coding.陪伴 vs 编码。
The key to performance is not just the prompt, but what data, tools, and context you provide to the model before the call. This is a new way of building software, which will require its own set of formal methods and infrastructure.性能的关键不仅仅是提示，而是在调用之前你向模型提供了什么数据、工具和上下文。这是一种新的软件构建方法，将需要一套全新的形式化方法和基础设施。
AI is a massive productivity boost. It democratizes creation, leading to more software and more people building it, not less.AI是巨大的生产力提升工具。它使创造民主化，将带来更多的软件和更多的构建者，而不是更少。
It's not because it's hard to build. People buy software because it defines workflows, operational logic, and how to use data. The hard problem is product design, not coding.不是因为它难构建。人们购买软件是因为它定义了工作流程、操作逻辑以及如何使用数据。难题在于产品设计，而不是编码。
Median lines of code in a PR: 2PR中代码行数的中位数: 2