No Clear Winner in AI Coding
Zach Lloyd, CEO of Warp, on the 3 phases of AI coding
The real opening is that AI coding still has no settled control point. The winning surface has already shifted from autocomplete inside the editor to agent workflows in terminals, chat panes, and background automation, which means distribution, product shape, and monetization are still moving. That keeps room open for an independent tool like Warp, especially because it sits at the command layer where agents can touch code, Git, Docker, CI, and external systems in one place.
-
Cursor and Windsurf proved there is huge demand, but they did not lock the market. Cursor jumped from $100M ARR at the end of 2024 to $1.2B by the end of 2025, while Windsurf was sold to Cognition after reaching $82M ARR in July 2025. Fast growth happened before any stable category winner emerged.
-
Warp is competing from a different workflow. Instead of living mainly in the code editor, it starts from the terminal, where developers already run tests, inspect logs, deploy code, manage Docker, and call CLI tools. That makes Warp a natural place for interactive agents today and triggered agents later.
-
Independence matters because model labs and incumbents want to own developer distribution. Warp is multimodel rather than tied to one provider, and its team context layer, shared commands, MCP setups, notebooks, and environment knowledge can make switching painful in a way raw model access alone does not.
The next phase favors products that become the operating layer for agent work, not just a smarter text box. If Warp keeps turning terminal activity, team context, and external tool integrations into programmable agent workflows, independence becomes an advantage, because it can aggregate the whole stack instead of defending one model, one editor fork, or one narrow feature.