Lou Bichard

Lou Bichard is a platform engineer and field CTO at Ona, the autonomous software-development company formerly known as Gitpod. He builds the infrastructure that lets coding agents work across enterprise repositories without sacrificing security, oversight or product quality.

Earlier, Bichard was principal engineer for developer experience at DAZN, where he worked on the streaming company’s internal developer platform. He founded Open Up The Cloud, an educational project focused on cloud-computing careers, and created a public platform-as-a-product resource collection. At Gitpod, he moved between product management and field-CTO responsibilities, examining how Kubernetes configuration drift disrupts developer feedback and how identity and access management can integrate organizational credentials, repository permissions and cloud infrastructure.

His central architectural contribution is making secure development environments practical for regulated enterprises. Gitpod evolved from multitenant Kubernetes and cumbersome self-hosted deployments toward customer-controlled cloud infrastructure: lightweight ECS runners create isolated EC2 environments backed by EBS storage inside customers’ AWS accounts, keeping source code and other sensitive assets within established security boundaries. The design reduces operational overhead while preserving auditability, and the same environments can support both human developers and autonomous agents. Bichard details these tradeoffs in his account of enterprise agent-fleet architecture.

Gitpod became Ona in September 2025, repositioning its development infrastructure around autonomous software engineering. Bichard’s guide to parallelizing Claude Code distinguishes individual coding agents from the isolated compute, persistent state and orchestration required to operate them reliably at scale.

• Swarms, fleets and event-driven agents: Swarms divide one objective among cooperating agents; fleets distribute work across repositories; event-driven agents respond to tickets, pull requests or schedules. Fleet execution suits vulnerability remediation, test-coverage improvements and other repetitive changes spanning large organizations.

• Virtual-machine isolation: Bichard favors fully provisioned virtual machines for enterprise coding agents because shared containers can introduce weaker security boundaries, compute contention and noisy-neighbor problems.

• Harness engineering: Repository instructions, tests, architecture notes and acceptance criteria encode organizational expectations directly into an agent’s working environment. When agents repeat predictable mistakes, improving those instructions and feedback loops can prevent future failures.

• Agent coordination: Bichard considers runtimes, orchestration and triggers comparatively mature; the unresolved challenge is transferring work between agents, managing limited context, enforcing development stages and introducing meaningful compliance gates. His proposed direction combines explicit workflows, durable execution and agent-accessible command-line controls.

• Software factories: With Zacharias Malguitou, Bichard built Memo through a ten-day autonomous-development experiment, reporting 375 merged pull requests, more than 67,000 lines of code and 1,067 tests without human-written production code. The experiment also exposed persistent weaknesses: imprecise specifications, automated checks that missed usability problems, and the continuing importance of human product judgment.

OpenAI completed its acquisition of Ona on August 10, 2026, bringing the company’s customer-controlled execution infrastructure into the Codex organization. Bichard’s individual position following the acquisition has not been publicly specified.