Solomon Hykes created Docker, the open-source container platform that transformed how developers package and run applications, and co-founded Dagger, which makes software-delivery workflows programmable and gives autonomous coding agents isolated environments. His work has repeatedly turned difficult infrastructure problems into portable, reusable developer tools.
Hykes founded dotCloud in Paris with Kamel Founadi and Sébastien Pahl, initially serving as chief executive. Its internal container technology became Docker, which he first demonstrated publicly at PyCon in 2013. As the company reorganized around that project, he became chief technology officer and helped establish shared container-infrastructure foundations, including runC, containerd, and the Moby Project. He left Docker’s daily operations in 2018.
With co-founders including Sam Alba and Andrea Luzzardi, Hykes subsequently built Dagger around programmable software delivery: builds, tests, and deployment workflows expressed as maintainable code and executed in portable containers. Dagger supplies reusable modules, language SDKs, caching, execution traces, and infrastructure primitives spanning filesystems, repositories, secrets, and network services.
Giving coding agents workable environments
Hykes extends these infrastructure principles to autonomous development through Container Use, an open-source system for giving coding agents isolated, configurable workspaces. Its design reflects several specific convictions:
- Complete environments, not isolated commands. Agents need their own files, tools, services, and execution settings throughout development; sandboxing individual commands does not adequately prevent shared-workspace conflicts.
- Git-backed agent workflows. Ephemeral containers and Git-backed snapshots preserve agent state while allowing developers to inspect changes, review history, enter an agent’s environment, merge useful results, or discard failed experiments.
- Parallel execution with human control. Multiple agents should work asynchronously in separate environments while developers retain practical opportunities to intervene, instead of approving every action or waiting passively for a finished pull request.
- Model-independent infrastructure. Coding agents, models, execution infrastructure, and secret-management systems should remain interchangeable. Model Context Protocol integrations let external agents such as Claude Code and Goose operate within defined project rules and scoped permissions.
At AI Engineer World’s Fair 2025, Hykes demonstrated containerized agent development, remote execution, Git-based inspection, and concurrent coding experiments, then made Container Use public onstage. His focus is making autonomous software development portable and observable while keeping infrastructure choices—and intervention—under human control.
