Warp CEO Zach Lloyd introduces Warp 2.0: AI agents that run in the terminal and why model competition drives quality

Summary

Warp launched version 2.0 on June 24, 2025, marking a deliberate pivot from a modernized terminal tool to a full agentic development environment. Zach Lloyd, founder and CEO, positions the product as an interface where developers prompt agents to handle coding, debugging, and production issues directly within a terminal-style UI. Warp claims its coding agent ranks toward the top of the SWE-bench eval and holds the number one position on the Terminal Bench eval.

Lloyd's thesis on the near-term developer workflow is straightforward: prompt-first replaces hands-on. Rather than opening files and typing commands manually, engineers will increasingly dispatch multiple agents in parallel, one fixing a bug, another resolving a production incident, with the developer acting as a supervisor who steps in when needed. He frames the engineer's role as one of orchestration and multi-threading rather than direct code authorship.

On model strategy, Warp currently defaults to Anthropic's Claude models, specifically Claude 4, Sonnet, and Opus, citing benchmark performance and user preference. Lloyd also notes competitive interest from OpenAI and Google's Gemini, the latter notable for its long context window. His stated preference is for a fragmented, competitive model provider market, because consolidation around a single provider would compress quality improvement and push up inference costs for application-layer companies like Warp.

The Anthropic dependency carries strategic risk Lloyd acknowledges openly. Warp is simultaneously Anthropic's customer and a competitor in the developer tooling space Anthropic has signaled it wants to monetize at scale, potentially tens of billions of dollars annually. Lloyd draws a parallel to the AWS dynamic, a platform that both provides infrastructure and competes with builders on top of it, and says the relationship is currently collaborative. He stops short of calling open-source model parity a prerequisite for business durability, but frames multi-provider optionality as a hedge worth preserving. A single-provider world, he notes, is a structural risk for any application-layer company, not just Warp.