Warp Bets on Open Source with GPT-5.5

Startup Warp is making a significant wager on the future of software development, betting that AI agents, powered by OpenAI's latest models like GPT-5.5, can drive open-source collaboration. The company is integrating GPT-5.5 into its 'Open Agentic Development' framework, aiming to automate coding tasks while humans focus on defining objectives and verifying results.

This approach positions AI agents to handle planning, coding, testing, and pull request generation, a shift from traditional development workflows. Warp’s CEO, Zach Lloyd, highlighted that GPT-5.5 enables agents to reason across larger problem spaces and has shown a 30% reduction in token usage compared to GPT-5.4 for agentic coding tasks, boosting efficiency for long-running workflows.

Warp, which began as a high-performance terminal, has evolved into a platform for AI-native development. The company open-sourced its terminal client and introduced Open Agentic Development, a model designed for building software collaboratively in the open. This model relies on AI agents for execution and humans for oversight and strategic direction.

Agent Orchestration with Oz

At the core of Warp’s strategy is Oz, its cloud orchestration platform. Oz manages AI agents across local and cloud environments, providing a control plane for deployment and coordination. Developers can initiate agents via a web interface, select skills, choose models, and monitor workflows.

Oz facilitates persistent, parallelized agent operations by incorporating shared memory, reproducible environments, and evaluation systems. OpenAI models are crucial within Oz, classifying tasks and routing complex coding and reasoning work to advanced configurations like GPT-5.5. These models also serve as 'LLM-as-a-judge' systems in Warp’s internal evaluation pipelines.

The Future of Collaborative Coding

Warp believes that effective agent orchestration can lead to more consistent code output than loosely coordinated human teams. In this paradigm, open source evolves from direct code contributions to human input on product judgment and shared vision.

The company's internal benchmarks show AI agents are now involved in approximately 90% of Warp’s pull requests, providing invaluable insights into the needs of scaling agent workflows, such as observability and human review. This mirrors the broader trend of AI agent workflow tools becoming essential for modern development.

Warp's commitment to open development extends to community involvement, allowing developers to help shape the evolution of orchestration and supervision systems as agents gain autonomy. This collaborative approach is seen as vital for navigating the uncharted territory of agentic development, a field where OpenAI models for coding are playing an increasingly prominent role.

The company reports significant growth, with its ARR increasing 35x last year, driven by organizations seeking scalable agent solutions. This rapid adoption underscores the market's readiness for advanced agentic development frameworks like Open Agentic Development.