Zach Lloyd, founder and CEO of Warp, used a keynote at the AI Engineer World’s Fair to outline the company’s move from a modern command-line tool to a broader platform for automated software development. Warp’s new orchestration product, Oz, is presented as a way to run continuous, automated development loops that triage, implement, review, verify, ship and monitor code changes.
From Rust terminal to factory concept
Warp launched in mid-2021 as a Rust-based modern terminal, before the ChatGPT era. As AI capabilities advanced, Warp added integrated coding agents to the terminal. In recent years the market for CLI tools has become more crowded, with competitors such as Claude Code, Codex CLI and Gemini CLI backed by large tech firms. In April of this year Warp open-sourced its core CLI tool.
Oz: agent orchestration and workflow integration
Oz is Warp’s response to what Lloyd describes as an industry transition from interactive agent use toward automated systems that continuously drive development. Oz aims to connect multiple models and coding harnesses across local environments and isolated cloud sandboxes while integrating into developers’ existing workflows.
When users enter Oz, they will be guided to set up a factory: choose repositories, decide which parts of the software lifecycle to automate, and specify where humans must remain in the loop. The platform can pull issues from Jira or Linear, accept submissions through Slack or Microsoft Teams, and allow developers to redirect agents via GitHub. Lloyd stressed that much of the factory’s value is not a new UI, but the integrations that let automation sit inside established workflows.
The terminal remains important
Lloyd says Warp’s mission hasn’t changed: help developers ship better software faster. The terminal remains an important product — Warp is improving it with community contributions and agent-driven processes. Warp placed its open-source CLI repository under Oz’s control and built a factory around the project, using Warp itself as a test bed for the factory approach. At the same time, the terminal is used by nearly a million developers and remains actively developed by Warp engineers.
‘Factory engineering’ as a new discipline
While the term “factory” may sound mechanized to individual engineers, Lloyd argues it should be viewed as a new engineering discipline: meta-engineering that builds the systems which in turn build software. Practitioners will ask why one agent succeeds and another fails, how to adjust feedback, which context an agent needs, and how workflows should change. He believes these skills will be intellectually interesting and necessary as automation scales.
Role of forward-deployed engineers and deployment models
Setting up a software factory often requires deep integration with existing systems and context, so forward-deployed engineering roles — people who can configure and deploy these systems inside companies — are likely to be important. Warp positions itself primarily as a platform business rather than a services provider, but Lloyd acknowledges there is currently demand for specialists who help organizations transform workflows with these products.
Gradual adoption, not instant replacement
Lloyd expects adoption to be incremental. Companies will start with low-risk repositories or specific use cases, measure performance, and gradually increase the share of automated merges and processes. Some work will likely remain human-led because of complexity or the need for creative thinking, but Lloyd predicts that within the next year every significant software project will have some continuous automation — an engine resembling a factory — driving it forward. He expects this approach to become as standard as GitHub or CI/CD in serious projects.
Practical advice for AI engineers
At the conference Lloyd urged AI engineers to experiment: instead of building features manually, try automating an annoying part of your workflow and build the loop. For example, attempt to make an agent implement incoming user issues automatically and observe what blocks adoption — code-review bottlenecks, lack of visibility into changes, or integration gaps. These experiments reveal the real obstacles to productionizing automation.
Scale and timing
Warp’s CLI is used by nearly one million developers. The core CLI was open-sourced in April, and Oz is being introduced as the company’s platform for the next phase. Lloyd believes the shift toward automated development loops will accelerate over the coming year, with factories becoming a standard element of how significant projects operate.
In short, Warp’s repositioning responds to rapid AI advances and heightened competition by building a platform that automates large parts of the software lifecycle while integrating with existing developer tools and workflows. Lloyd sees “software factories” as a practical, widespread evolution of how software will be produced.
