Most write-ups about AI and software development stop at the prompt. Someone shows you how they got Claude to scaffold a component, the demo looks clean, and the post ends there. What nobody documents is what happens when that component ships and something adjacent breaks at 2pm on a Tuesday.
I've been digging into how AI-assisted dev teams actually operate in production, and this breakdown on EngineeredAI.net is one of the clearest field reports I've seen: https://engineeredai.net/ai-human-dev-team-workflow/
The framing that clicked for me: vibe coding is solo. Vibe team software engineering is coordinated. Same underlying tooling, completely different operational model.
The team structure described is four roles. Founder acting as PM, an AI dev agent handling implementation, a QA engineer as the human verification gate, and a QA collaborator managing the context and documentation layer. That last role is the one most people underestimate. The AI agent doesn't carry institutional knowledge between sessions. Someone has to make that context explicit, or the team loses coherence fast.
The loop itself is tight: ticket filed, agent picks it up, fix shipped, QA verifies, ticket closed. On a well-scoped ticket with a clear reproduction path, that closes same session. That turnaround does not exist in a traditional dev team where a bug report might sit in backlog for a week.
But the seams are real. The agent can't move its own tickets. Speed creates mutations — a fix to one surface breaks something adjacent, and the agent only tested what it was asked to fix. Context gaps between sessions require documentation discipline that most solo operators skip until it bites them. And loose ticket scope causes agent drift: technically correct fixes that break the spirit of the requirement.
The practical takeaway for anyone building on top of AI agents: the quality of the human coordination layer is the actual constraint. Tight tickets, explicit scope, a human verification pass that checks for intent drift not just acceptance criteria. That's what separates a functioning AI dev team from a pile of AI-generated rework.
The roles that survive this shift are the ones that require judgment, verification, and institutional memory. Implementation work that can be clearly specified is already at risk. The job isn't gone — the shape of it changed.
Top comments (0)