The software inflection point: agents, OSS, and InnerSource

Agentic development has shifted the bottleneck in software from writing code to defining what code to write. Drawing on examples from Naval Ravikant, Warp, Spotify, Anthropic, and OpenAI, the post argues that open source and InnerSource projects need to become 'agent-ready' — with AGENTS.md files, repo-specific skills, MCP servers, and encoded scope/architecture gates — to safely receive agentic contributions. The maintainer's role evolves from writing code to defending vision and reviewing at scale. For InnerSource specifically, agentic SDLC could dissolve the social friction that has historically prevented cross-team contributions, since the host team retains control over specs, gates, and merges. Key risks include review capacity becoming the new bottleneck, the migration burden falling on already-stretched unpaid maintainers, spec quality replacing code quality as the critical variable, and questions about whether 'open-sourced' repos with proprietary orchestrators are truly open source.