Summary
Agent 4's central innovation is a canvas interface that serves as a 'cockpit control panel' for an entire business: users can generate designs, build backends, create automations, develop mobile apps, and build slide decks in one interface. Under the hood, every task forks a complete virtual machine and database into an isolated environment where an agent runs autonomously, enabling parallel background work while the user focuses on creative and design tasks. Masad frames this as a return to Replit's REPL (read-eval-print loop) philosophy after agentic coding temporarily pushed UX toward a mainframe-style 'type and wait' pattern.
On the business: Masad says Replit has grown strongly organically, particularly in enterprise — naming Mercedes, PayPal, Plaid, and Axos as customers, spanning auto, fintech, and financial services. The $400M raise (investors not named; leaks attributed to investor community gossip) gives Replit capital to compete in paid marketing as the space becomes more competitive. Masad says Replit doesn't optimize for ARR metrics because in the AI era, ARR can be gamed by burning tokens without delivering value. Instead, they track sentiment (agent anger signals), successful deployments, and increasingly, revenue generated by businesses built on Replit via Stripe integrations. On hiring: Masad says new grad hiring is accelerating because young people are 'agent maxing' — running 100 parallel agents — which is exactly the skill profile Replit wants.