system of record

The talk describes a shift in YC startups toward hard tech, with growth in robotics, industrial manufacturing, defense, semiconductors, photonics, and power. Speakers attribute this to stronger technical founders, advances in AI-assisted research and coding, and renewed investor interest; they also share examples from space, defense, and compute infrastructure. On the software side, they argue that agents doing entire workflows can deliver more value than point solutions or systems that only track work, while data and reinforcement learning environments are becoming businesses in their own right. They also discuss the rise of solo founders and experienced founders, and advise builders to start experimenting with current AI models and coding agents.

The talk argues that enterprise AI projects often fail because internal teams and consultants struggle to build useful software, while startups can combine technical skill with product judgment and an understanding of business processes. It explains why selling AI to large organizations takes time: startups must integrate deeply with existing systems, navigate internal politics, and earn the trust of a champion. Examples from Taktile, Greenlight, Castle AI, and Reduct show startups winning work after internal or incumbent solutions fell short. The speakers encourage engineers to try AI coding tools on real projects and see the difficulty of implementation as an opportunity to build AI-native software.