Warp’s big bet on building open source with GPT-5.5

Warp, a modern terminal provider used by nearly 1 million developers and over 56% of the Fortune 500, is scaling its Open Agentic Development framework using OpenAI models, including GPT-5.5. According to the company, humans define objectives and supervise outcomes while agents plan work, write code, test changes, and open pull requests. Frontier AI improvements have made this orchestration practical, with GPT-5.5 using 30% fewer tokens per task than GPT-5.4 in internal benchmarks and helping agents reason across larger problem spaces. Within Warp, agents now co-create about 90% of internal pull requests. To manage these workflows, Warp built Oz, a cloud orchestration platform that acts as a control plane for deploying and coordinating agents across local and cloud environments. Oz incorporates techniques like context compaction, persistent memory, and dedicated subagents to maintain reliability across extended workflows. OpenAI models are utilized for demanding agentic coding tasks and as LLM-as-a-judge systems within evaluation pipelines. Warp's open-source terminal client launch featured OpenAI as the founding sponsor, inviting the developer community to help shape the evolution of orchestration and supervision systems. This agent-centric shift coincides with significant financial growth, as Warp reported a 35x increase in ARR last year and enterprise revenue up over 500% since Q4 2025, driven largely by organizations seeking flexible ways to scale agent workflows.