Leadership Garden

AI

How AI is changing engineering work, team design, and the quality bar leaders need to maintain.

11 posts. Read them in order; each post links to the next.

Read in order

  1. No, AI didn’t break PRs

    AI didn’t break pull requests; it exposed the review queues, slow integration, and weak collaboration habits engineering teams already had.

  2. AI Is an Amplifier. What Are You Amplifying?

    AI doesn’t care about quality. It accelerates whatever direction your team was already heading. Here’s what engineering leaders need to change.

  3. The Work Moved: What the AI Coding Debate Actually Agrees On

    AI agents generate 98% more PRs but reviews take 91% longer. The work didn’t disappear — it moved. A synthesis of eight perspectives on where it actually went.

  4. Verification Debt Is Your Next Headache

    Verification debt is accumulating quietly in AI-assisted codebases. What engineering leaders need to watch for.

  5. A hard job attached to an easy idea

    Most engineering orgs run a hard job attached to an easy idea. How to tell the difference, and what to do about it.

  6. The Same Feature, Twice

    The same feature took a weekend on a laptop and a quarter in the product. The model was identical both times. Everything around it was not.

  7. Specs Don’t Drive Anything

    Spec-driven development’s defenders describe a loop. Its diagrams draw a pipeline. An engineering leader on where the spec actually belongs.

  8. The Cost of an Objection Went to Zero

    AI lets you produce work you can’t evaluate. Worse, it lets everyone else critique yours. What engineering leaders should actually do about it.

  9. Everyone Has a Recipe Now

    AI lowered the barrier to building software, not to understanding it. Why spec-first AI recipes look right in week one and fail by month nine.

  10. You’ll Know in Two Quarters

    Why “just ship faster” stops working when feedback takes months, and what engineering directors should measure instead.

  11. AI Is Not Just A Tool

    “Just a tool” lets AI rollouts skip the hard question: what will your engineers stop learning?