A year ago, developers debated whether AI could write production code. That debate is over.

GitHub Copilot is now used by 77% of Fortune 500 companies. Cursor crossed $2 Billion in ARR. The developer who writes every line manually is already a minority. AI handles the boilerplate, the refactoring, the test coverage. The human handles architecture, judgment, and direction.

That is one layer of the shift. It is large. It is not the whole story.

The deeper shift is happening one layer below.

A new category of developer is entering the market. They do not open a terminal. They do not configure a local environment. They do not write syntax. They describe what they want in plain language, and the software appears. They iterate on the description until it works. Then they deploy.

Replit has over 50 million users building this way. Lovable is growing 50% month over month. These are not prototypes. They are production platforms with compounding user bases.

Three forces made this possible at the same time.

First, AI became capable enough to translate natural language into working, deployable code. Not toy code. Production code. The threshold crossed quietly, and then it was just true.

Second, browser-based infrastructure matured to support full-stack development without local setup. No installation. No dependency management. No environment configuration. The entire stack lives in the browser. Figma proved this model in design. Notion proved it in documents. Replit is proving it in software.

Third, the next generation of developers is entering the market through this interface, not despite it. They are learning on Chromebooks in Southeast Asia, tablets in bootcamps, phones in Lagos. For them, describing intent and getting working software is not a shortcut. It is the default. Traditional coding is the foreign workflow.

These are not competing products. A developer can build a prototype in Replit, export the code, and refine it in Cursor. The layers are complementary. The question is which platform owns the first touch with the next generation of creators, because first touch compounds into habit.

The two layers of the market are also distinct in what they require from a platform. Cursor and Copilot serve developers who already know how to code. They accelerate an existing workflow. The moat is integration depth and model quality. Replit and Lovable serve people who are creating software for the first time through description. The moat is the full stack: intent interface, compute, database, deployment, community, all in one place. You never need to leave. The more you build there, the harder it is to move.

Network effects compound that. The larger the user base, the more templates accumulate, the more community solutions exist, the higher the switching cost for the next user.

The company that consolidates intent-based development in the next 18 months will control how the next billion people create software. That market is worth $50B to $100B at maturity. Cursor owns the developer productivity layer. Nobody has locked up the creation layer yet.

Replit is the closest thing to that consolidation bet. Full-stack browser infrastructure. Enterprise integrations with GitHub, GitLab, Bitbucket, and Databricks that let users learn through intent as students and stay as they move into professional roles. It is a platform play, not a tool play.

We are evaluating it as such.

Bashar Aboudaoud
Managing Member, UpRound