Browse latest
Tools & PlatformsOpenAI News · May 27, 2026

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

Warp, a modern terminal, is betting on "Open Agentic Development," a model where humans define objectives and supervise outcomes while AI agents handle coding tasks. This approach, powered by GPT-5.5, aims to revolutionize software development by enabling agents to collaboratively write and test code.

Author: Morein.ai Editorial

Warp, initially a modern terminal, gained popularity among developers for its speed, collaboration features, and AI-native interface. Recognizing the growing role of coding agents, Warp envisioned the terminal as a natural hub for developers to interact with these agents, integrating commands, context, and collaboration.

This year, Warp open-sourced its terminal client, with OpenAI as a founding sponsor, and introduced "Open Agentic Development." This model empowers humans to set objectives and supervise outcomes, while AI agents plan, write, test code, and manage pull requests. This approach leverages recent advancements in frontier AI models, making large-scale agent orchestration practical.

GPT-5.5 significantly enhances Warp's open-source workflows by enabling agents to tackle complex problem spaces and prepare work efficiently for human review. Internal benchmarks show GPT-5.5 uses 30% fewer tokens per coding task than GPT-5.4, boosting efficiency in long-running agent workflows.

Warp's internal engineering demonstrates the model's success, with agents co-creating approximately 90% of the company's pull requests. This firsthand experience highlights the critical needs for scaling agent workflows: observability, coordination, memory, and human oversight. Warp has nearly 1 million developers and is used by over 56% of the Fortune 500.

To manage agents across local and cloud environments, Warp developed Oz, its cloud orchestration platform. Oz acts as a central control plane, allowing developers to deploy and coordinate agents, select skills, choose models, and monitor workflows. It also offers features like context compaction, persistent memory, and dedicated subagents to maintain agent reliability in extended workflows.

OpenAI models play a crucial role within Oz, classifying tasks and routing complex coding and reasoning to stronger model configurations, including GPT-5.5. Warp also uses OpenAI models as LLM-as-a-judge systems in its evaluation pipelines. Warp's "Open Agentic Development" and the Oz platform represent a long-term vision where software development evolves from individual interactions to coordinated systems of persistent agents, a bet that has already resulted in significant growth for the company.

Read original source

Related articles