Jon Peck is a senior product marketing manager at Evinced working to make accessibility a practical requirement of AI-assisted software development. Previously, he helped build GitHub’s enterprise advocacy practice and taught organizations how to adopt coding agents without surrendering engineering standards.
A software developer since the late 1990s, Peck studied computer science at Cornell University and worked with Cornell, Massachusetts General Hospital, and Algorithmia. At Algorithmia, he focused on serverless machine-learning infrastructure, making machine-learning capabilities easier for developers to discover, deploy, and combine.
He subsequently spent nearly six years at GitHub in customer engagement, developer relations, and enterprise advocacy, helping establish its Enterprise Advocacy team in the Americas. His writing covered collaboration-ready repositories, choosing AI models for practical coding tasks, and turning well-scoped issues into agent-generated pull requests.
- Start with existing software. Peck favors brownfield AI adoption: applying coding assistants to established applications through bounded tasks such as migration, test generation, documentation, and infrastructure configuration. He measures usefulness by engineering outcomes, not the volume of generated code.
- Give agents institutional knowledge. Detailed READMEs, scoped issues, shared knowledge bases, and repository-level Copilot instructions encode architectural decisions, security expectations, and team conventions. Those instructions should evolve whenever engineering standards change.
- Restrict what agents can change. Peck’s PostgreSQL integration demonstration used read-only Model Context Protocol integrations to retrieve database information and generate realistic test fixtures without modifying the database. Agent-created branches and pull requests remain subject to human review, security checks, and existing CI/CD controls.
