Which Side of the Build Gap Are You On? (Run the Scorecard)

This is the first post in a nine-part field-guide companion to Builder-Leader: The AI Exoskeleton That Crosses the Gap. One post per chapter, each one a thing you actually do, not a thing you read. The full exercises live in the book's public repo; these posts are the standalone version.

Start with the one that needs no tools and takes fifteen minutes: figuring out where you actually stand.

Three gaps, stacked

When people argue about what AI can do, they're usually standing on different ground without knowing it. There isn't one gap between you and a calibrated read on AI capability. There are three, stacked.

The tier gap. If the most advanced AI you've used is the free tier of ChatGPT, Copilot in Office, or Apple Intelligence, your mental model is calibrated to models one to two years behind the frontier. You're right about what you've seen. What you've seen is old.

The distribution gap. AI advances fastest in domains with a clean success signal: code that compiles, a test that passes, a proof that checks. It advances slowest in open-ended writing and advice. If your direct experience is mostly drafting and summarizing, you've been watching the slow lane and reasonably concluded the car is slow.

The build gap. This is the one the book is about, and the one the scorecard measures. It's the distance between using these systems and building the structure around them.side quest

The build-gap scorecard

Rate each, 0 (never) to 4 (multiple times this quarter). Be honest about the last six months, not your plans.

1. I've personally written a skill, a rule, a CLAUDE.md entry, or any file that instructs an AI system how to behave for my work.
2. I've watched an agent work for more than thirty minutes on a task I gave it and intervened only to course-correct, not to take over.
3. I've debugged a failed agent run by reading the logs or the agent's reasoning, not by asking someone what went wrong.
4. I've written a hook or a guardrail that stops an AI system from doing something I don't want it to do.
5. I've added a Model Context Protocol (MCP) server to my own setup, or removed one deliberately.
6. I've composed multiple agents or subagents to work in parallel on independent pieces of a task.
7. I've looked at a peer's setup and could tell whether it was better than mine, or worse, and why.
8. I've explained to a direct report what an agent harness is and been confident in the answer.

Add it up.

• 0-8: You're on the far side of the build gap from the operators the book describes. In April 2026 this is the modal position for VP-and-above, and it's not a judgment. It's who the book was written for.
• 9-16: You've touched the work but haven't moved past individual experiments. The harness isn't yet something you carry.
• 17-24: Beginning operator. The book accelerates you past the first two months of the six-month ladder.
• 25-32: You're past the build gap. The book is still useful for the vocabulary and the arguments you'll need to make to peers who are where you were eighteen months ago.

The questions aren't arbitrary. Each one names a specific capability that separates someone who uses these tools from someone who operates them: instructing behavior, supervising autonomy, debugging from reasoning, setting guardrails, managing tools, composing agents, evaluating setups, and explaining the structure. If you scored low, those eight are a curriculum, not a verdict.

The sentence that does the real work

The scorecard tells you where you are. This tells you why.

Finish this in writing: "The specific reason I have not personally operated an agentic AI system more than I have is..."

Some answers are reasonable. No time. No board pressure. My team's output with these tools is already good enough. The tools aren't mature for my domain.

Some answers aren't. I pay someone else to do it. I evaluated it once in 2024 and it wasn't ready. I delegated to my CIO. My assistant handles it.

Read your sentence back. If you cringe, you've found the right question, and the rest of the field guide is the answer to it.

The one commitment

If you scored under 17, there's exactly one thing to do before you go further: install Claude Code on your own machine and run the first hands-on exercise (that's Part 6 of this series). Put it on a calendar this week.side quest

You don't have to buy the larger argument yet. You only have to cross the install threshold once. Everything past the diagnosis assumes you've done it.

The full Chapter 1 field guide, with the tier and distribution scoring, is in the public repo: field-guide/01-two-groups.md. The book is at builder-leader.com.

Next in the series: hearing the three tiers in a real conversation, and naming the failure mode your org is actually stuck in.