Driven
Humans drive session-bound agents from the IDE, CLI, shell, or chat loop.
Agent-native engineering
Agents help immediately. They compound when the lifecycle adapts.
witwave is a cloud-native autonomous agent framework for AI agent teams and a live project using that framework to maintain itself: Kubernetes-native runtime, MCP tools, durable identity, shared context, memory, observability, human escalation, and compounding improvement.
Delegated
Agents take bounded tasks but still deliver into the old tickets, PRs, and review queues.
Native
Agents become governed participants in the software development lifecycle itself.
Start here
Two foundational papers
These papers anchor the site. One explains the adoption path for agentic AI in software engineering; the other explains the AI agent team architecture that becomes possible once agents are part of the lifecycle.
Whitepaper
The Three Phases of Agentic AI Adoption in Software Engineering
Driven -> Delegated -> Native. A framework for understanding how agents create value in existing lifecycles, and what changes when teams integrate them into the SDLC.
Read the paperWhitepaper
Anatomy of an Agentic Team
A practical model for specialized AI agents, coordination, memory, process improvement, and the human role in an agent-native operating model.
Read the paperWorking model
Meet the agents behind the operating model.
The public roster makes the thesis tangible: a manager, builders, reviewers, a documenter, release captain, and outreach lead, all working as a governed agent team rather than a collection of disconnected chat sessions.
Meet the team
Field notes
Field notes, essays, and launch posts live here.
The blog collects field notes, release reflections, and project essays, with links to public conversations around each piece when they exist.
Read the blog