You can go further than you expect.

Where are you?
And how far do you want to go?

Add your answer in the live sheet:

Answer in the google sheet

Keypoint: Bounded contexts unlock parallel work.

• 4–6 streams — one bounded context each.
• Reach files + tools — CLI, Copilot, or client apps.
• Map yours in the sheet — tools you use + one gap.

Write your stack

Keypoint: Fewer handoffs — room to run more workstreams in parallel.

• Less middleware — AI creates, you check.
• AI connects to your files and approved tools.
• Steer more streams by cutting steps, not typing faster.

Keypoint: When AI runs execution, you can plan and steer more workstreams at once.

How ready are you to shift from doing the work to guiding AI that does it?

Answer in the live sheet:

Answer in the google sheet

Keypoint: One bounded space per engagement — AI stays inside; swap tools, not context.

• Files or tenant spaces — AI works inside your boundary.
• Context stays put when you change model or vendor.
• PM briefs and dev repos — one engagement, one folder.

LLMs go in many directions. Your job is to focus the model.

• What to ignore and what to prioritize.
• Name who reviews before it ships.
• A strong prompt becomes a skill.

Keypoint: Stable workflow → save as skill → wire into a chain.

• Skill: saved instructions the model reads each run.
• Chain: skills in order with gates (plan → build → judge).
• When work changes, update the skill — keep the chain as the map.

Keypoint: One prompt starts a chain — skills work, personas gate, then ship.

"Update the skills slide — teach us the chain while you ship it live."

• Personas review before any slide file changes.

Keypoint: Lock intent before you plan — one question at a time.

".grillme: Grill this slide — one question, your best guess."

• Best guess each time — you confirm or correct.
• Locks outcome, scope, and success criteria.
• Output: intake file, then .plan.

Keypoint: Give AI a clear goal — it loops until done while you steer and review.

• /goal keeps Codex working until the objective is met.
• You pause, redirect, or approve — same pattern as PM skills.
• Good goals name success criteria, limits, and proof.

Keypoint: shape problems clearly — agents build; you judge what ships.

• Translation layer compresses — spec is becoming the product.
• Problem shaping, context curation, and taste — core PM skills.
• Bottleneck upstream: what’s worth building.

Recommendation: find verbs you repeat — skills, loops, and your approval before anything ships.

• Any process with a repeated verb is a skill candidate.
• Loop until your test passes: “Would I send this as-is?”

Recommendation: use AI when change needs a fast, consistent update.

• Skill: turn rough notes into a stakeholder brief.
• Loop: detect change → draft update → you approve.
• Works for projects, delivery, support, and leadership.

Recommendation: start reviews with a summary — you verify before you comment or merge.

• Skill: PR diff summary plus risk flags for reviewers.
• Skill: test failure triage from logs and stack traces.

Recommendation: shipping is a verb too — draft the boring parts, you own the final wording.

• Skill: release notes from commits or changelog entries.
• Loop: collect → draft → you trim → publish.