Interview, Fireside Chat
Building the Generative Web with AI ft Vercel CEO Guillermo Rauch
- Vercel aims to scale its guidance to a user base ranging from single-digit to double-digit millions of developers, anticipating a "year over year" doubling of users as AI opens the top of the funnel.
- The company predicts a "one in a generation" shift toward a "generative web" where applications are created on demand for individuals, potentially rendering traditional downloadable software obsolete within an unwinnable battle.
- A two-class developer split is forecasted: one group creating apps via conversational interfaces (captured by the "v0" side of the world) and another group creating platforms requiring traditional tools, with LLMs eventually producing code "secure by construction."
- Over the "next three years," companies failing to adopt AI-native models face the collapse of existing "kingdoms," while new AI-native entities are expected to rise to "unprecedented heights" rapidly.
- The "next five years" are projected to bring the "biggest transformation" to the web regarding interfaces, user habits, and creator participation, driven by an "LLM-first" publishing model where users ask questions rather than performing keyword searches.
- Infrastructure plans include "automatic scaling for any kind of workload" as a best-case scenario and autonomous investigation of problems as a worst-case outcome, with AI agents replacing manual performance recommendations.
- Vercel's infrastructure is expected to prevent "tens of thousands" of vulnerabilities by stopping LLMs from leaking secrets, citing a current rate of "1,000 vulnerabilities prevented per day" due to AI's ability to identify flaws human engineers lack the bandwidth to find.
- Future business structures will evolve through a lifecycle of no AI, AI prototyping, production-grade AI teams, and full business transformation, leading to a future where "all software will be AI" and "all money will be digital."
- The outlook includes the emergence of "virtual designers, marketers, and developers" as expert agents, alongside "QA agents" that continuously check critical paths, while "agents that automatically do the performance optimizations" become standard.
- Dynamic systems will require "robots that are constantly watching everything and repairing it" as agents become their own modality, with a long-term focus on "asynchronous agents" capable of prolonged collaboration with humans.
- The web is viewed as the ultimate platform for hosting AGI, with expectations that personal or ephemeral apps will become so seamless users will not notice they are ephemeral, while "whatever breaks the norm goes from zero to taking over the world" once norms are perfected.