Software 1.0 (SaaS) was a rigid, static form of intelligence. Some smart, highly motivated group of software-savvy humans get together, try to spot patterns and workflows in a given domain, smash some Red Bulls, stand up a database, wrap some UI around it, bundle it into a binary, and hit publish. Buyers are happy they’re finally understood and cared for by a software vendor. “This company Boopitify from Palo Alto finally cracked the problem!” As the product and market matures, more customers join the platform, the initial simplicity erodes little by little. The platform slows down by the endless stream of features, bug fixes, dialogs with nested dialogs, integrations, permission systems, Enterprise™️ requirements. The platform eventually gets so generic, it gets dumber, and it makes you, the user, feel dumb and disempowered. AI cuts right to the chase. You’re no longer buying a rigid proxy for intelligence. You’re buying ingelligence itself, measured in tokens. This is, incidentally, the same the cloud did for infrastructure. No more rigid, static, fixed-size data center. Everything is consumed on-demand, and it’s on tap. The key enabler of this has been code generation. Whether it’s vibe coding, GenUI, or agents, the software we used to code, is now generated. The interfaces we used to design: generated. The prototypes we used to spec: generated. For vendors, it means rethinking your interfaces. Agents or assistants rather than rigid UIs. MCPs rather than rigid integrations. For AI labs, it means winning the coding war. Agents need to write and execute their own code to solve problems. For infra providers, it means rethinking what the “hello world” app is. It’s increasingly not web services. This is why we’re building the AI Cloud, made up of AI services, for AI apps, orchestrated by AI agents.