Vibe Coding for the Next Billion | Michele Catasta, Replit | RAISE Summit 2026

Replit Strategy and Market Trajectory

• Michele Zanini, President and Head of AI at Replit, initiated a 15-year research journey to democratize software creation, transitioning from open-source contributions to deep learning research at Lausanne and Stanford.
• Replit launched its first agentic workflows in September 2024, roughly six months before Andrej Karpathy coined the term "vibe coding," prioritizing early market feedback over perfection.
• Replit has surpassed 60 million users and $500 million in implied revenue, targeting $1 billion in revenue by the end of the year.
• The company reports exponential growth in customer success stories, explicitly shifting focus from "token maxing" metrics to long-term business revolution via software automation.

User Demographics and Organizational Impact

• Replit targets "knowledge workers" and "citizen developers"—non-technical employees who require software solutions to automate departmental tasks in HR, go-to-market, and support functions.
• The platform adopts a general-purpose agent strategy rather than targeting a single persona, allowing adoption across diverse organizational units.
• Michele predicts a workforce shift by 2025 where every employee becomes an "agent manager," responsible for prioritizing tasks and vetting results from autonomous agents rather than executing steps manually.
• Early adopters of this agent-centric workflow report productivity increases significantly higher than those from six months prior.

Technical Architecture and Model Strategy

• Replit intentionally avoided investing heavily in training its own foundational models, recognizing that capital-intensive frontier research would distract from building infrastructure and agent harnesses.
• The current model strategy is heterogeneous, combining fine-tuned proprietary models, open-weight models, and frontier models based on specific task requirements and complexity.
• Replit has introduced "VibeBench," a new benchmark specifically designed to evaluate web coding capabilities of large language models.
• The company relies on the rapid pace of open-weight model innovation, noting that these models are becoming impressively capable for web coding tasks.

Future Product Vision and Market Outlook

• The long-term vision involves a shift from building static applications to building autonomous agents that can execute end-to-end goals without human UI interaction.
• Future iterations will allow users to define high-level business goals (e.g., "increase conversion rate by 5%") and delegate the execution of experiments, copy changes, and design adjustments entirely to agents.
• Michele argues that while some job displacement is inevitable, the primary impact will be "upgrading" human roles by removing tedious tasks, enabling a shift toward strategic thinking and new business creation.
• Leaner companies are expected to emerge as the cost of labor decreases through the automation of tedious work via AI agents.

Call to Action and Adoption

• Users are urged to embrace AI technology immediately, even in early or imperfect stages, drawing parallels to the utility of early chatbots which saw widespread adoption despite limitations.
• Agent penetration remains below 1% globally, contrasting with over a billion users for chatbots, representing a significant gap in market adoption and potential impact.
• Michele emphasizes that technology moves faster than previous technical revolutions, advising builders to stay ahead by testing rough prototypes that will likely mature within six months.