🧠 Human Skills

John Ousterhout distinguishes two ways of programming: tactical and strategic.

It gives you the vocabulary to see where AI is already better than you.

Two modes; reference book; map of dev work.

Day to day: writing code, syntax, finding bugs, creating commits.

It’s exactly the work AI does more cheaply and better.

Win the battle; execution; syntax and bugs.

Think long term: what the codebase should look like and how to work faster.

It’s where your advantage lies — the one AI still can’t replace.

Win the war; architecture; speed.

AI has "eaten" tactical programming—it executes tasks more cheaply and better.

Insisting on competing tactically means swimming against the tide.

Commoditization of tactical work; an infinite fleet of AI.

You need to excel at strategy to get the most out of your tactical AI fleet.

Someone who manages the fleet well is worth far more than someone who only executes.

Move from tactical to strategic; command the fleet.

Spend your time on architecture, speed, and roadmap — not syntax.

It’s the practical application: redirecting your hours toward strategic work.

General at the top; architecture, speed, roadmap.

"Your skills are the ceiling for what AI can do."

Shows that AI is a multiplier, not a replacement for your expertise.

Multiplier; the skill's own ceiling.

AI makes senior professionals about 10x better at what they already do.

The bigger your foundation, the more AI can unlock.

Gain proportional to the baseline; senior 10x.

Juniors get only a small boost; hiring many of them isn't worth it.

Reframe how you think about teams and where to invest.

Little benefit at the beginner level; a team strategy.

People with low skill can’t get the best out of AI.

Your lack of expertise becomes the ceiling that limits the tool.

Low ceiling; the human becomes the bottleneck.

Getting good with AI = getting good in your domain. A better teacher uses AI to teach better.

Investing in yourself is the highest-leverage investment.

Domain = leverage; a better teacher with AI.

Improving yourself means improving AI—because the multiplier acts on your foundation.

It gives you a direct application: study and practice what actually raises the ceiling.

Raise the ceiling; multiplier on the base.

Knowledge is the fundamental understanding you carry in your head.

It’s the foundation, but by itself it doesn’t guarantee you’ll know how to act.

Understanding; knowing what and why.

Skill comes from doing something many times — muscle memory.

Distinguishes “knowing about” from “knowing how to do.”

Repetition; muscle memory; practice.

Wisdom is knowing WHEN to do something and how it fits into the real world.

It’s what separates the competent from the truly good.

Timing; judgment; fit with reality.

Wisdom is almost impossible without living the exact context — to be like someone at Anthropic, you'd need to go to Anthropic.

Explains why you can’t “download” wisdom from outside.

Lived context; situated wisdom.

Knowledge and skill can be packaged into reusable skills.

It’s what lets you distribute expertise to AI and the team.

Skills package knowledge + ability.

Wisdom can't be packaged — it remains your irreplaceable role.

Defines where you need to stay in control, always.

Delegation limits; wisdom that can't be packaged.

You design the difficult parts up front, before delegating.

The hard decisions are exactly what AI shouldn’t make on its own.

Difficult decision made early; upfront design.

Define the scope of what will be done very carefully before assigning the task.

Vague scope is the #1 cause of incorrect results.

Clear scope; task limits.

Think about the interfaces between modules—the contracts that connect the parts.

Good interfaces let AI work in parallel without breaking everything.

Contracts; module boundaries.

Think through test scenarios and write good tests.

Tests are the safety net that lets AI implement things safely.

Scenarios; safety net; good tests.

Sufficient documentation that points AI to the right places, and a codebase that’s easy to work with.

The right doc reduces back-and-forth and speeds up delivery.

Directional docs; navigable codebase.

The practices are the same as delegating to a junior/mid-level developer.

You already have the mental model; you just need to apply it to AI.

Classic delegation; management discipline.

Being able to put your vision into words, out loud, for the AI.

AI only executes well what you can articulate.

Articulate the vision; think out loud.

The saying is “overpowered”—you put out much more content, much faster.

Speaking is much faster than typing; broadband changes the game.

Input speed; overpowered dictation.

"The speed at which you move tokens out of your brain and back into it."

Seeing the cycle helps you optimize the right bottleneck: communication.

Brain↔token cycle; bandwidth.

Communicating and speaking are ridiculously overpowered in today's dev world.

Communication has become one of the most valuable technical skills.

Communication as a technical skill.

Matt uses Whisper Flow to dictate; dictation tools put this into practice.

The right tool turns speech into real input.

Whisper Flow; dictation in the workflow.

Speaking well is a skill you can train — not a fixed gift.

If it can be trained, it’s worth practicing deliberately.

Trainable fluency; deliberate practice.

AI is notoriously bad at coming up with original ideas.

Shows that the product's originality is still yours.

AI's limitations; human creativity.

"You should be in charge of the product" — you choose the features, have the vision and know the why.

Defines the irreplaceable role of the product owner.

The product command; the vision and the why.

Ask the AI what to REMOVE, how to simplify and improve the UX—not "what's the next big feature."

Use AI where it helps (critique, simplification) and not where it falls short (vision).

Remove > add; simplify the UX.

The fundamentals of product design haven’t changed; the books are still relevant.

You don't need to relearn product thinking—just apply it with AI.

Durable product fundamentals.

Talking to customers, finding out what they need, and building prototypes — that stays the same.

Discovering a real need is human and irreplaceable.

Customer discovery; prototypes.

The fundamentals of business haven’t changed; AI only gives you an execution advantage.

Set the right expectation: AI speeds things up; it doesn’t replace strategy.

AI handles execution; strategy remains human.