The a16z Show

a16z General Partner Martin Casado sits down with Grepr CEO Jad Naous to ask what comes after AI learns to write code: can it also debug and operate the systems that code creates? Jad shares how AI has transformed engineering inside his 10-person team, cutting some projects from six months to just 10 days. But operating production systems is a harder challenge, with unpredictable workloads, failures, and countless possible states. They discuss whether AI can learn to diagnose problems, test hypotheses, and debug systems like a human engineer. They also explore AI SREs, security, query optimization, and why the best engineers may be gaining the most leverage from coding agents.

a16z General Partner Alex Rampell sits down with VGS Executive Board Member, Terry Angelos, to unpack what happens to payments when AI agents go from recommending what to buy to actually making the purchase. Terry explains the infrastructure required to safely give agents access to payment credentials, why wallets like PayPal, Shop Pay, and Link are competing to become the default payment layer for agents, and how new agentic tokens could preserve the chargebacks, identity checks, and merchant relationships that make today's payment networks work. VGS, which already secures payment credentials across a significant share of e-commerce, is now provisioning millions of these agentic tokens each week. They also discuss what happens when merchants don't want to transact with agents, how delegated authority changes liability when an agent makes a mistake, and why AI could simultaneously increase consumer spending while eliminating the subscriptions, fees, and other forms of economic "breakage" that depend on people not paying attention.

a16z Partner Lisha Li sits down with AI researcher and Trillium founder Nathan Lambert to ask how durable the gap really is between frontier AI labs and open models. Nathan argues that OpenAI and Anthropic have structural advantages in compute, capital, and the ability to pioneer new training paradigms, but open-model labs, particularly in China, have proven remarkably capable of following close behind. They unpack how much of that comes from distillation versus technical talent, why compute remains a major constraint, and how data, RL environments, and post-training are increasingly shaping model progress. They also discuss Nathan’s new nonprofit research lab, Trillium, and its plans to open up parts of AI research that are increasingly difficult to study from the outside. That includes releasing RL environments and documenting how AI is actually being used to automate AI research, with the goal of demystifying recursive self-improvement and giving researchers and policymakers a clearer picture of what frontier systems can and cannot do.

Could AI break the cryptography protecting everything from crypto wallets and banking to private messages? Recent advances in AI-driven mathematical research have sparked new warnings about the security of today's encryption systems, but how much of the concern is supported by evidence? a16z crypto's Liz Harkavy is joined by Stanford cryptographer Dan Boneh, a16z crypto General Partner Eddy Lazzarin, and Research Partner Justin Thaler to separate what we know from what remains speculation. They explain why modern cryptography relies on mathematical assumptions, what OpenAI's recent results do and don't demonstrate, and how the potential threat from AI differs from quantum computing. They also discuss why existing security practices and cryptographic alternatives offer reasons for confidence, the importance of building systems that can adapt if an assumption fails, and how AI could ultimately help researchers and developers make software more secure.

a16z General Partner Martin Casado sits down with MakerMaker.AI co-founders Dhaval Adjodah and Owen He to unpack recursive self-improvement, or RSI, and separate the increasingly popular term from what it would actually mean for an AI system to improve itself. Dhaval and Owen distinguish true RSI from simply using AI to automate pieces of AI development. Their goal is a closed loop where the system being improved is also doing the improving, across components including models, harnesses, and eventually more of the AI stack. They explain why closing that loop still doesn’t guarantee runaway “takeoff,” and why measuring real progress requires accounting for both capability gains and the resources required to achieve them. They also share what they’re seeing in their own research, including early signs that self-improving systems can increase performance while improving efficiency. And they dig into the practical challenges of proving those gains are real, from variance and reward hacking to the enormous difficulty of distinguishing genuine improvement from noise.