Most important take away

AI didn’t create a meetings problem — it exposed a team size problem. The optimal unit is a 5-person “strike team” because AI raised per-person output by 10x, which also raised the coordination cost of each additional person by 10x. The correct response isn’t to shrink headcount but to restructure into small teams and 10x your mission ambition.

Chapter Summaries

The Meetings Problem Is a Team Size Problem Meetings have tripled since 2020. AI note-taking apps don’t fix this — they amplify output through a broken coordination structure. The root cause is teams that are 3-10x too large.

The Math of Team Size (The Number Five) Communication pathways between people grow exponentially: 5 people = 10 pathways, 10 people = 45, 20 people = 190. Dunbar’s research (1992), military testing, Bezos’s two-pizza rule, and Fred Brooks’s 1975 software engineering findings all converge on ~5 as the optimal team size for deep coordination.

What AI Actually Changed AI raised per-person output 5-10x (AI-native companies run $2-3M revenue per employee vs. traditional SaaS at ~$250-500K). This means the coordination cost of adding a 6th person is now measured in millions of lost productivity, not thousands.

Correctness Over Volume AI made volume free. The scarce resource is correctness. A Harvard Business School study (2025) at Procter & Gamble found AI-augmented teams were 3x more likely to produce top-10% quality ideas — not 3x more output, but 3x more likely to be right.

Two Archetypes: Scouts and Strike Teams Scouts operate solo with full AI toolkit for exploration (example: Peter Steinberger building OpenClaw in 60 days). Strike Teams are 5 people focused on execution where correctness matters — every person’s output passes through at least one other brain with shared context.

The Ambition Expansion (Not Cost Reduction) A 500-person company with AI didn’t get a cost reduction — they got the capacity of 2,500-5,000 people. The question isn’t “how small can we get” but “what becomes possible when every 5-person team has the capacity of a department?” Example: 400 engineers restructured into 80 strike teams could build a 10-product platform.

Hiring and Talent Identification Give people scout missions — a real problem, full AI tooling, one week, clear objective. Test for: Can they define problems without specs? Do they know what “right” looks like architecturally? Do they default to action or permission? Your most frustrating employees (who skip meetings and build without asking) may be your best strike team members.

Summary

Actionable Insights:

• Restructure teams to ~5 people. This is the biological and mathematical optimum. Above 5, coordination costs now destroy value at an unprecedented rate due to AI-amplified output per person.

• Optimize for correctness, not volume. AI made volume free. The teams that win will be the ones that ship things that are actually right — architecturally sound, strategically coherent, and free of subtle errors.

• Reframe AI as a force multiplier, not a cost cutter. Instead of asking “how many people can we cut?” ask “what mission was impossible before that’s now achievable with our current people reorganized into strike teams?”

• Run scout missions to identify talent. Give individuals a real problem, full AI tools, and one week. Look for people who can define problems without specs, hold systems in their head, and default to action over permission.

Career Advice:

• Develop “taste” and architectural judgment — the ability to know what “right” looks like at the system level. This is the most valuable skill in the AI era, not syntax-level coding.
• Become a generalist architect who uses AI to extend into adjacent domains rather than a narrow specialist in one lane.
• Stop clinging to coordination skills (running meetings, writing status updates, navigating org structures) as your primary value. These are overhead in strike teams.
• Embrace scout-mode behavior: build things without asking, define problems yourself, ship without committees. The organizational structure that punished this is becoming obsolete.
• Toby Lutke’s Shopify mandate: prototype with AI before any real build, demonstrate why AI can’t do a task before requesting headcount, make AI fluency part of performance reviews. This is the direction leadership is heading.