Keep pulling the thread on Replit Agent.
In the past 6 months, engineers at Replit have nearly tripled their code output.
From early January to late June, Replit experienced a 5.8x increase in lines of code contributed by its engineers.
Controlling for hiring, a consistent cohort of authors at Replit produced 2.9 times as much code as before the implementation of internal agentic systems.
Replit tripled its per-engineer code output rate while simultaneously doubling the size of its engineering team.
Replit's agent-assisted code review process has saved 30% of human pull request review time, and this figure is growing.
Replit's AI team built a continual learning system that allows the Replit agent to be self-improving by analyzing user feedback, proposing improvements, and using benchmarks and A/B tests for validation.
Replit's support team closes the most difficult tickets, those escalated to humans, 60% faster by using an internal AI agent.
Replit plans to productize its internal 'self-driving' company technology and offer it to all its users.
A fundamental prerequisite for creating a 'self-driving' company is integrating AI agents with all existing operational systems across the organization.
Pulsia, which began as a framework for companies with no human employees, has evolved into an AI-native operating system designed to minimize the effort required to build and maintain a company.
Despite a near-tripling of code output at Replit, code review times have held steady, while reversions and product incidents have remained flat.
AI agents at Replit are used to investigate production incidents, review pull requests, answer questions, analyze business data, triage support tickets, and research sales accounts.