概要
公開日: 2026-05-27
翻訳生成に失敗したため、原文をそのまま保存しています。
原文
May 27, 2026 Startup Warp’s big bet on building open source with GPT‑5.5 Warp uses GPT‑5.5 to orchestrate agents across local, cloud, and open-source workflows. Start building with OpenAI (opens in a new window) Company size: Startup Region: North America Industry: Technology Products: API 30% Fewer tokens per task with GPT-5.5 90% Of internal pull requests created with agents Loading… Share Warp (opens in a new window) started as a modern terminal, earning early love from developers for its speed, collaboration features, command workflows, and AI-native interface. As coding agents moved from experiments to everyday engineering workflows, Warp saw the terminal becoming a natural place for developers to work with agents: where commands, context, collaboration, and review already meet. When Warp open-sourced (opens in a new window) its terminal client this year, with OpenAI as the founding sponsor of the repo, the company also introduced Open Agentic Development: a model for building software in the open. Humans define objectives and supervise outcomes, while agents plan work, write code, test changes, and open pull requests. Recent improvements in frontier AI models helped make that kind of agent orchestration practical at scale. For Warp’s open-source workflows, GPT‑5.5 helps agents reason across larger problem spaces and prepare work for human review. In internal benchmarks, GPT‑5.5 used 30% fewer tokens per agentic coding task than GPT‑5.4, helping Warp improve efficiency as it scales long-running agent workflows. Today, Warp has nearly 1 million developers and is used by more than 56% of the Fortune 500. In Warp’s own engineering organization, agents now co-create around 90% of the company’s pull requests, giving the team a firsthand view into what long-running agent workflows need to scale: observability, coordination, memory, and human review. “We think we can ship a better Warp, more quickly, by working with our community to supervise a fleet of agents. OpenAI models help make that sustainable for the long-horizon coding work these systems require.” —Zach Lloyd, CEO The next generation of collaborative software development Open Agentic Development is Warp’s bet on where software development is heading. Agents will write code, and developers will specify intent, verify the outputs, and decide what ultimately ships. Those choices become reusable context for future agents, allowing the system to improve over time. If the orchestration is good enough, Warp believes agents can produce more consistent code than a loosely coordinated group of humans. Open source then becomes less about humans contributing implementation work directly, and more about contributing the product judgment and shared vision that only humans can provide. Persistent, parallelized agents need components like shared memory, reproducible environments, evaluation systems, permissions, and ways to coordinate work. Warp built Oz, its cloud orchestration platform, to manage agents across local and cloud environments. For Open Agentic Development workflows, Warp uses GPT‑5.5 for agents that help manage open-source contributions, according to the company. OpenAI models have also performed strongly in Warp’s internal evaluations for long-horizon engineering tasks involving reasoning, planning, code generation, and code review. Agentic orchestration with Oz Oz (opens in a new window) acts as a control plane for deploying and coordinating agents across local and cloud environments. Developers can launch agents through a web interface, select predefined skills and environments, choose the model and hosting configurations, and monitor long-running workflows centrally as they execute. Once launched, agents can continue running remotely while developers inspect live sessions, monitor execution state, review generated artifacts, and hand workflows back and forth between cloud and local environments without losing context. Oz also supports recurring workflows, allo