Sequoia Capital US partner Konstantine Buhler recently delivered a speech titled "The $10 Trillion AI Revolution," which is well worth a watch. Key points: 1. Industry Stage Assessment The Industrial Revolution took 211 years to refine the "steam engine → factory → assembly line" model; The AI revolution has only progressed to "GPU steam engine → deep learning factory." The next phase will be "dedicated assembly lines"—specific AI systems for vertical scenarios will replace today's general-purpose large models and become the birthplace of new monopolies. 2. Market Size Reassessment The AI service market is not $20 billion, but $10 trillion (Sequoia Capital simply adds "number of displaced occupations x annual salary"). SaaS in the cloud era has boosted the $350 billion software market to $650 billion; the AI era will recalculate all current "human services," pushing their boundaries outward. 3. Five major trends determine the strategy. 1) Productivity leverage: Hundredfold efficiency trades for decreased controllability → Products must make "human-machine verification" the default workflow. 2) Evaluation criteria move beyond the lab: Real business metrics > academic benchmarks, starting with a scenario-based "competition." 3) Practical reinforcement learning: RL is no longer just a paper, but a core formula. Whoever first develops "RL + vertical data" will win. 4) AI enters the physical world: With the software dividend exhausted, hardware manufacturing, quality inspection, logistics, and robotics are the next 10x growth areas. 5) New computing power production function: FLOPs/person increases by 10-10,000 times. Inference efficiency and security represent a hidden trillion-dollar market. 4. Sequoia's Five Current Areas of Investment 1) Persistent Memory (Long-Term Memory + Identity Consistency) 2) Seamless Communication Protocol (the "TCP/IP" of AI-to-AI) 3) AI Voice (Faster than video, most lucrative in B2B scheduling/transaction scenarios) 4) AI Security (Three-layer protection: R&D, distribution, and endpoints, enabling standardized "security agent" products) 5) Open Source Competitiveness (Preventing ecosystem lock-in by giants, continuously investing in open source models and toolchains) 5. One-Sentence Action List for Entrepreneurs Step 1: Choose a vertical human resources track with "annual salary x headcount > $50 billion"; Step 2: Use RL + proprietary data to train a "small model" and first achieve world-leading performance on a real-world KPI; Step 3: Integrate "persistent memory + multi-agent collaboration" into the product, allowing users to directly hire an AI team; Step 4: Convert the 50% cost savings into a new pay-per-result business model; Step 5: In 2018 Feed business data back into the model within a month, creating a data flywheel and achieving a 10x efficiency difference over competitors—that's your "dedicated assembly line." 15 Keywords Specialization, 10 trillion replacement costs, RL engineering, AI physicalization, FLOPs production function, persistent memory, AI voice portal, secure agent, open source moat, real KPI competition, data flywheel, small vertical models, human-machine verification workflow, pay-per-results, efficiency difference x10. Speech link: youtube.com/watch?v=yoycgOMq1t...…