AI + a16z

Patrick Collison, CEO of Stripe and lifelong programmer, reflects on his Smalltalk and Lisp roots. He recounts Stripe’s early Ruby and MongoDB choices and the long work of API design. He talks about using LLMs for research and coding. He explores AI’s role in productivity and ambitious efforts to train models that ‘program’ biology.

Guido Appenzeller, a systems-focused product thinker; Joel de la Garza, a security and risk leader; and Yoko Li, an open-source builder of personal AI assistants. They dig into OpenClaw integrations and long-running agent use cases. They talk about a seven-hour Gmail setup, agents requesting broad permissions, why consumer sites resist agent-friendly interfaces, and strategies for identity, authorization, and containment.

Vishal Misra, Columbia professor and vice dean specializing in how LLMs function, explains experiments that show transformers update token predictions in mathematically predictable ways. He compares LLMs to human learning, highlights the plasticity gap, and discusses why moving from pattern matching to causal, continual learning is vital for true AGI.

Amjad Masad, CEO and co-founder of Replit — creator of a browser-based coding and AI platform democratizing software development. He talks about building millions in revenue, how AI shifts coding toward product and UX, why he turned down a $1B offer, practical automations that create value, and why AI empowers builders rather than ends them.

Martin Casado, General Partner at a16z and former networking/VMware engineer-operator, discusses VC specialization, the rising importance of AI infrastructure, and why media and distribution matter for firms. Conversation touches on talent competition, the role of open source, and how coding and developer productivity are evolving with AI.

Sarah Wang, a16z general partner focused on AI/model investing, and Martin Casado, a16z general partner expert in infrastructure and networks, discuss the unique capital dynamics reshaping AI. They cover why fundraising now directly drives capability, the blurring of apps and infrastructure, the risk of model firms out-raising their ecosystem, talent wars, custom hardware economics, and two divergent industry futures.

Samar Abbas, CEO of Temporal and creator of the durable-execution system that began as Cadence at Uber. He explains why long-running AI agents need persistent state and recoverability. He describes how Temporal powers production workloads like OpenAI Codex and Snap, and why background, multi-agent systems create new distributed systems challenges at internet scale.

Ankur Goyal, founder and CEO of Braintrust and former databases/AI products engineer, joins to talk engineering, evals, and productionizing models. He explains what evals do and why feedback loops matter. They debate systems vs. scaling mindsets, compare SQL vs. Bash agent designs, and unpack open vs. closed model cycles and Chinese model dynamics.

Sam Altman, CEO of OpenAI and entrepreneur-investor in AI and energy, discusses Sora as a research and societal preview, the case for vertical integration and massive infrastructure, next-gen interfaces beyond chat, AI-driven scientific discovery, safety and regulation for superhuman models, and why energy (solar, storage, nuclear) matters for AI scale.

Martin Casado, general partner at Andreessen Horowitz who invests in AI infrastructure, joins to explain why today’s AI surge is not the dot-com collapse. He breaks down where AI dollars go: data-center GPUs, power, and cooling. He outlines infrastructure layers a16z backs and points to opportunity in the long tail of generative-AI beyond flagship LLMs.