The Engineering of Intent, Chapter 12: The GenDD Pod

This is Part 12 of a series walking through my book The Engineering of Intent. In the previous chapter, we covered agentic debugging. Part V of the book zooms out from the individual to the team. Chapter 12 introduces the team structure that actually ships AI-native work sustainably.


Three People. Continuous Flow. Almost No Ceremony.

The GenDD Pod has three members, continuous flow, and almost no ceremony. It is not a reorganization trick. It is a recognition that when one engineer plus one agent does the work of three, the team shape that produced five-to-seven-person squads no longer makes sense.


The Three Roles

  • Agentic Product Lead — owns the Specs. Translates business intent into architecture.md, conventions.md, and the intent documents the agents consume.
  • Agentic Engineer — owns the Model/Tool interface. Prompts, Context Packs, agent orchestration, skill design.
  • Agentic QA Engineer — owns the gate stack. Custom rules, tuned tests, the E2E harness, the reviewer agent.
💡 Why three: Three covers all four pillars (Vibes, Specs, Skills, Agents) without constant context switching. Four adds communication overhead; two starves QA. Three is the steady state. The Product Lead and Engineer collaborate on Vibes and Specs. The Engineer and QA collaborate on Skills and Agents. The Product Lead and QA collaborate on outcome measurement. Everyone stays close to the customer.

Transitioning From Agile (Without the Caricature)

The transition that works: one team, one month without mandatory ceremonies, explicit permission to experiment, measurable goals. Observe what forms organically, codify what works, expand from there. Designing the target state top-down before collecting evidence reliably produces a caricature of GenDD that nobody can execute.

And measuring the pod honestly matters. Lines of code is misleading. The right metrics are outcome-shaped: features shipped, customer problems solved, incident rate, cycle time. End-of-month one-page “what we shipped and why it mattered” summaries are the best defense against velocity theater I’ve found.

“A mid-sized SaaS went from seventy-two engineers to twenty-four over eighteen months. Feature throughput was essentially unchanged; incident rate dropped; morale was mixed. The transition is hard on people. Do not run this play without a human-cost plan.”


Anti-Patterns and Hiring

The anti-patterns I keep seeing:

  • The rockstar pod that hoards interesting work.
  • The parallel-ceremony anti-pattern — running Agile alongside GenDD because nobody wants to kill the old rituals.
  • The QA-as-gatekeeper framing that turns QA into the brake instead of the accelerator.
âš  Hiring notes: Product Lead is the hardest role to hire for — interview for writing, not presentation. Engineer: look for shipping output, not years of experience. QA Engineer: look for platform-builder instincts, not manual-testing pedigree. The shape of these roles has changed more in the last two years than in the prior ten, and the old interview loops will lead you wrong.

Next up — Chapter 13: VibeOps and CI/CD Evolution. The pod needs a platform. Chapter 13 is about how the deployment, observability, and rollback surfaces change when the code flowing through them was co-authored with an agent — and the operational habits that separate teams that ship confidently from teams that ship nervously.


📖 Want the full picture?

The chapter walks the three-role decomposition in depth, the Agile-to-GenDD transition playbook, the honest-metrics toolkit, the twelve-squads-to-four-pods case study with the full 72-to-24-engineer transition, the complete anti-pattern catalog, and the interview rubrics for each of the three pod roles.

Get The Engineering of Intent on Amazon →

2026-04-28

Sho Shimoda

I share and organize what I’ve learned and experienced.