State of Startups and AI 2025 - Sarah Guo, Conviction

Too Long; Didn't Watch — Summary

Sarah Guo of Conviction highlights that AI is the largest technology revolution, driving rapid value creation and user adoption, with significant advancements in model capabilities and the rise of AI agents; she emphasizes that execution is the primary moat for startups, encouraging engineers to build "thick" AI applications by deeply understanding specific workflows and extending AI's impact beyond coding.

Main Takeaways

• AI is driving unprecedented user uptake and value creation, with companies achieving rapid ARR growth.
• AI capabilities are rapidly advancing in reasoning, agents, and multimodal applications, while inference costs are plummeting.
• The market for AI models is becoming highly competitive, with new players and open-source alternatives challenging incumbents.
• Successful AI applications, like those in coding, thrive by deeply understanding user workflows and redesigning them with AI, a strategy applicable to all industries.
• Execution, rather than first-mover advantage or proprietary models, is the critical differentiator and moat for AI startups.
• Co-pilots and augmentation are currently more effective than full automation due to human tolerance for error and latency.
• Significant opportunities exist in applying AI to complex, data-scarce problems in fields like robotics, biology, and material science.

Detailed Summary

Introduction (0:00 - 4:36)

• Sarah Guo introduces herself as a partner at Conviction, an AI-native venture fund started nearly three years ago, just before ChatGPT's launch.
• She states that AI represents the largest technology revolution, bringing immense change to technology, business models, and product decisions.
• Conviction invests in infrastructure, model, and application-level companies, including Cursor, Cognition, Mistral, Harvey, and Open Evidence.
• Guo notes the unprecedented user uptake in the last few years, despite AI product and engineering being harder than anticipated.
• She highlights massive value creation, with companies reaching $10-100 million ARR faster than in previous tech revolutions.
• The focus should be on long-term product viability and usage numbers, rather than market hype cycles.

AI Capabilities and Market Dynamics (4:36 - 11:15)

• Reasoning: Identified as a key capability, allowing AI to scale intelligence with more compute and unlock new use cases requiring transparent, high-stakes decisions and systematic search.
• Agents: Defined as software that plans, includes AI, takes ownership of tasks, holds goals in memory, and can try hypotheses and backtrack; the number of agent startups has increased by 50% in the last year.
• Multimodality: Progress in voice, video, and image generation is creating real businesses (e.g., HeyGen, ElevenLabs, Midjourney exceeding $50M ARR).
• Voice is expected to see initial business applications due to its natural communication mode, with other modalities following as they become more controllable and less costly.
• Capability improvement is expected across the model layer, with no