Nine Notes on Making Things.

Make to think.

We don't truly understand what we think until we've made it.

We can hold an opinion, defend it, argue it in a meeting. That's not the same as knowing anything. Knowing means we've put the thing in front of the world and watched what happened.

This is why I trust prototypes more than decks. A deck says what you plan to make. A prototype shows you what happens when the plan meets the actual world. Most plans don't survive that meeting.

This isn't new. called it inquiry a hundred years ago. You don't know a situation until you've acted into it and something has pushed back.

The bottleneck moves.

Things don't die. The cost just moves somewhere else.

When someone says “X is dead,” what they usually mean is “X used to be expensive and now it's cheap.” Those aren't the same thing. The cost of X hasn't gone away. It moved somewhere else.

AI made making things cheap. It didn't make any of the harder things cheap: choosing what to make, building a product people actually trust, getting it in front of the right people. Those got more valuable, exactly because the making got easy.

built a whole career on this. Every technology enhances something and obsolesces something at the same time. Look for what got scarce.

Judgment isn't a rubric.

There's an 85% version of almost anything. The last 15% is what makes it good, and it never gets written down.

Almost every craft has an 85% version. The part you can teach, the part you can write down, the part you could hand to a model and get a decent result back. That's what fills cookbooks and style guides and training programs and AI outputs. Real work, and it matters.

It's just not what makes the thing good. The last 15% is what makes it good, and nobody has ever managed to write it down. It lives in the person who's watched a thousand of these fail. wrote a book about this. said a cookbook can't make a cook.

This is why “AI will replace design” is a category confusion. When making gets cheap, judgment gets scarce, and judgment is what the last 15% is made of. Cookbook doesn't make a cook. Prompt doesn't make a designer. The bottleneck moves.

Choose your rooms.

Quality is felt directly. Value is what a system decides you're worth. They aren't the same thing.

Quality and value are two different things and people confuse them all the time. Quality is what happens when a person looks at your work and reacts to it directly. Value is what a system says you're worth, and systems (a company, a market, a promotion committee, the ranking algorithm on some aggregator) don't measure everything.

You can do extraordinary work and get paid nothing for it, because you're in a room that doesn't know how to see what you actually did. You can also do mediocre work and get rewarded well, because the room can see it clearly. called this legibility. Systems can only act on what they can measure, so they push everything toward measurable forms.

If your work isn't being recognized, the honest question is: is the work not good enough, or is this the wrong room? Both happen. Either can be fixed. The mistake is to keep trying to fix the work when the room was always the problem.

Fast to contact. Slow to judgment.

Sprint to make it. Then don't rush what it tells you.

Speed matters because until you make the thing, you don't actually know if the idea works. Every hour spent arguing about a design in the abstract is an hour not spent seeing what the design actually does.

But speed becomes a problem when it turns into a personality. If you're moving fast just to feel productive, you're not letting the thing you made teach you anything. You have to get there fast, then stop and really look at what you got.

This is basically 's two systems. System 1 is the fast instinct that gets you to the prototype. System 2 is the slow work of sitting with what you made. Good work needs both. The switch between them is the hard part.

Systems and people.

Systems raise the floor. People raise the ceiling.

Design systems, playbooks, org charts. The best ones absorb the lessons you've already learned, so you're not solving the same problem for the tenth time. They make everyone faster and they keep the floor high.

They also, eventually, become the ceiling. A system that never breaks is only showing you what it already knew how to see. called it the quality without a name. The specific rightness of a specific solution no rule could have predicted. If you don't leave room for the person who does something the system can't reward, all your team can do is ship.

Build the system. Then keep asking: what did this person figure out that isn't in the system yet?

The right to refuse.

Everyone deserves a voice. Not everyone has the same authority.

Collaboration improves decisions. It doesn't make them. Everyone in the room can bring useful input: evidence, constraints, taste, an objection you hadn't heard. That doesn't mean everyone should have equal say over what ships.

This is a question of ownership, not expertise. A designer can own the call without holding every kind of knowledge in the room. Accountability for the outcome is what creates the authority. It doesn't require exclusive claim on what's true.

If you're accountable for the outcome, you have to be able to say: I hear you, but not this. Remove that. We're not solving that problem right now. That feedback is real, but it doesn't belong in this version. Especially now, when AI can generate an infinite number of options, the person who says “no, not that” is doing the actual work.

wrote about interpretation as a kind of aggression. The way readers, critics, collaborators try to make the work mean what they already believe. The author's job is to hold the line: this, not that. That's not gatekeeping. That's the work.

The alternative is diffusion. No accountable owner, no point of view, work that drifts toward the average of what everyone in the room said. That gets called empowerment sometimes. It isn't.

Make more authors.

When the leader is the gate, the team learns to ship at 80%.

The failure mode of a strong creative leader is becoming the last quality gate. When the team knows you'll review everything and fix what's wrong at the end, they ship at 80% and let you close the gap. Blame the process, not the people.

Don't review harder. Give real ownership to more people. Not “you can weigh in.” Actual ownership of a real piece of the work, with the outcome on their name. called this power-with instead of power-over. Authority that flows from function, not position. She was writing in the 1920s. Most organizations still haven't figured it out.

The slogan version of empowerment is “you can share your opinion in the meeting.” The real version is “this is yours, ship it.” One of those you can put on a wall poster. The other one is what actually develops people.

Loose identity. Tight work.

Be strong enough to make a specific thing, and light enough to change your mind about it.

Being sure and being wrong aren't opposites. A point of view has to be strong enough to make you commit to a specific version of the thing, and light enough that you can change your mind when the evidence comes in. Be embarrassed by last year's work. That's a good sign.

But be exact about the work itself. Leave your hand on it. Refuse to make it look like nobody made it. made a useful distinction between work (the durable things you leave behind) and action (how you show up in front of other people). called the same self-directed move overcoming.

Most people do this backwards. They hold on tight to who they think they are (the job title, the discipline, the tribe) and let the work be whatever's asked. Do the opposite. Hold your identity loosely. Hold the work tight.

Nine things you can actually do about it.

Reading is easy. Changing what you do on Monday is harder. One move per belief, and when to use it.