AI software creation platform
Replit
Replit provides an AI software creation platform for individuals and teams, including people without programming experience. Replit Agent turns natural-language requests into applications, helping users create databases, connect interfaces to backend services, and deploy their work. Built-in authentication, hosting, monitoring and integrations support prototypes, internal tools and customer-facing applications. Its visual canvas lets users refine designs and apply them to an app, while collaborative workflows let teams coordinate tasks and run multiple agents in parallel.
Founded by Amjad Masad, Haya Odeh and Faris Masad, Replit incorporated in 2016 after the founders began working together in 2015. Amjad Masad remains CEO, and Haya Odeh leads design. The platform grew from a collaborative browser coding environment into an integrated app-building system. Its infrastructure work includes adopting Nix in 2021: configurable development environments draw dependencies from a shared package store, replacing the need to maintain one increasingly large operating-system image.
In March 2026, Replit reported over 50 million users, including users from 85% of Fortune 500 companies, and raised $400 million at a $9 billion valuation. Sacra estimated its annualized revenue run rate reached $525 million in April 2026. Replit offers free access alongside paid individual, professional and enterprise plans.
2 talks
Newest first
2 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- Building AI For All
Start here for Amjad Masad's account of programming's progression from punch cards to AI-enhanced development environments and the announcement of broadly available Replit coding assistance.
Amjad Masad · Michele CatastaAI Engineer Summit 2023
- The 3 Pillars of Autonomy – Michele Catasta, Replit
Learn how Michele Catasta distinguishes supervised from fully autonomous experiences and defines reducible runtime around execution without technical user decisions.
Michele CatastaAI Engineer Code 2025
Messages from the stage
Training smaller code models
In the 2023 session, Michele Catasta emphasizes training-data quality, repeated training passes, and HumanEval Pass@1 evaluation in explaining the case for smaller code language models.
Verifying extended agent execution
Michele Catasta's 2025 talk examines verification through static analysis, tests, API checks, and browser interactions, including directly generated Playwright code. It also addresses the difficulty of keeping long-running agent loops coherent.
Affiliations reflect each recorded session, not necessarily current employment.