New diff-tracking techniques in the coding arms race are introduced by Warp AI

Warp, the AI-powered coding platform, is pushing to make its coding agents more transparent and controllable—an approach that feels a lot like pair programming.

Warp Code, a set of new capabilities intended to provide developers with more control over command-line coding agents, is being introduced by the firm today. The update introduces advanced diff-tracking and clearer visibility into exactly what the agent is doing.

"According to CEO Zach Lloyd, "you are just crossing your fingers and hoping that whatever comes out on the other end is something you can actually incorporate. What we’re doing instead is creating a tight feedback loop around this agent-driven coding process.”

In practice, that means you can monitor the agent’s work in real time and even question its decisions along the way. “As the agent writes code, you’ll see every tiny change it makes,” says Lloyd. “You can comment on those differences and adjust the agent’s behavior on the spot.”

The interface will look familiar to existing Warp users: a command box at the bottom to direct the agent, a main window for responses, and a side panel that shows the agent’s step-by-step edits. Like code-centric tools such as Cursor, users can still edit code manually, but they can also highlight specific lines to provide context for requests. Perhaps the most impressive piece: Warp’s compiler automatically resolves errors whenever generated code fails to build.

Lloyd frames it this way: “The goal is to make sure you understand what the agent produces—and that you can edit and review it with confidence.”

This represents a fresh strategy in an increasingly crowded AI coding landscape. Warp competes not just with AI-powered editors like Cursor and Windsurf, but also with no-code platforms such as Lovable. Meanwhile, foundation model companies are offering their own command-line coding assistants, like Anthropic’s Claude Code and OpenAI’s Codex. (In fact, Warp itself relies on custom-built models to power its features.)

Warp is expanding quickly even though it is still a modest player in the AI coding race. With more than 600,000 active users, the platform is adding about $1 million in annual recurring revenue every 10 days—a sign that plenty of developers are willing to pay for a better way to code with AI.