Mike Krieger 探讨 AI 如何重塑产品开发，如今 90%的代码由 AI 编写，并分享了他对 AI 能力不断演变的看法。
Mike Krieger discusses how AI is reshaping product development, with 90% of code now AI-written, and shares his evolving views on AI capabilities.
要点 · TL;DR
AI 能力进步超预期，产品策略与用户理解变得比以往更重要。 AI capabilities are advancing faster than expected, making product strategy and user comprehension more critical than ever.
当 90%代码由 AI 编写时，瓶颈转移到决策、对齐和合并队列。 When 90% of code is AI-written, bottlenecks shift to decision-making, alignment, and merge queues.
MCP 是解决 AI 上下文与记忆的关键，可实现可组合、可脚本化的集成。 MCP is key to solving AI context and memory, enabling composable and scriptable integrations.
核心观点 · Key points
AI模型的发展速度超出预期，编码基准和智能体能力正按预测扩展。 AI models are advancing faster than expected, with coding benchmarks and agentic capabilities scaling as predicted.
产品团队应专注于策略、可理解性，并让用户看到AI的潜力。 Product teams should focus on strategy, comprehensibility, and opening users' eyes to AI's potential.
当90%的代码由AI编写时，瓶颈转移到决策、对齐和合并队列。 When 90% of code is AI-written, bottlenecks shift to decision-making, alignment, and merge queues.
MCP是解决上下文和记忆的关键，使AI集成可组合、可脚本化。 MCP is key to solving context and memory, enabling composable and scriptable AI integrations.
创始人应在模型能力边缘构建，拥抱破坏并迭代。 Founders should build at the edge of model capabilities, embracing breakage and iterating.
在AI驱动的未来，培养孩子的好奇心和独立思考至关重要。 Nurturing curiosity and independent thinking in kids is crucial for an AI-driven future.
反共识 · Contrarian takes
AI模型现在能有独立、创造性的观点，而不仅仅是平淡的回应。 AI models can now have independent, creative opinions, not just anodyne responses.
产品杠杆来自产品人员与研究人员合作，而不仅仅是构建用户体验。 Product leverage comes from embedding product people with researchers, not just building UX.
像ChatGPT那样的消费者心智份额不是唯一路径；专注于构建者可能更有价值。 Consumer mindshare like ChatGPT's isn't the only path; focusing on builders can be more valuable.
关闭Artifact是正确的；尽管被喜爱，但它缺乏增长和能量。 Shutting down Artifact was right; it lacked growth and energy despite being loved.
传统参与度指标对AI具有误导性；深度比频率更重要。 Traditional engagement metrics are misleading for AI; depth matters more than frequency.
AI应该是思想伙伴，而不是拐杖；保留用户自主权是关键。 AI should be a thought partner, not a crutch; preserving user agency is key.
本期章节 · Chapters（共 31）
引言Introduction
AI能力认知转变Changed Mind on AI Capabilities
AI的独立思考与创造力Independent thinking and creativity in AI
时间线预测与AI 2027Timeline predictions and AI 2027
避免可怕的AI情景Avoiding the scary AI scenario
培养孩子的好奇心Nurturing curiosity in kids
面向未来的技能Skills for the Future
AI编写代码与产品开发AI-Written Code and Product Development
审视Claude Code与合并队列Reviewing Claude Code and the Merge Queue
Patient Zero与Claude代码占比Patient Zero and Percentage of Claude Code Written by Claude
瓶颈与Claude的战略角色Bottlenecks and Claude's Role in Strategy
Kevin Weil与产品人谈模型团队Panel with Kevin Weil and Product People on Model Teams