We accidentally made an AI platform: Jamie Turner

Convex aims to replace traditional backend engineering by extending the functional reactive paradigm of frontend frameworks like React into the server. By pervasively tracking data flow and dependencies, the platform eliminates the need for manual cache invalidation, polling, and custom push mechanisms, allowing developers to focus on product features rather than infrastructure glue. This architecture has inadvertently become a primary choice for generative AI startups, which now comprise over 90% of projects on the platform. These applications leverage Convex to manage sophisticated, concurrent background workflows—such as speech recognition, summarization, and embedding generation—while seamlessly syncing results back to the user interface. To support this shift, the platform has integrated native vector indexing, launched a dedicated startup program, and introduced high-level state machine components to accelerate the development of complex AI-driven backends.