Every piece of work I have been proud of started as something embarrassing. A page of half-sentences. A layout with the wrong number of columns. A plan that fell apart the moment I said it out loud to someone else.

For years I read that stage as evidence I was bad at the job. Now I think it is the job.

## Writing is how the thinking happens

There is a comfortable myth that we think first and record second, that the work exists fully formed somewhere behind the eyes and the only difficulty is transcription. It has never once worked that way for me. I do not know what I believe about a problem until I have tried to say it plainly, and trying to say it plainly is where I discover that half of what I believed was borrowed, vague, or simply wrong.

So the first draft is not a bad version of the final one. It is a different object with a different purpose. It exists to find the argument.

> Write to find out what you know. Edit to make it worth someone else's time.

## Lower the stakes on purpose

The practical fix is small and slightly silly: make the first pass somewhere that does not feel precious. A plain text file. The back of an envelope. A message to yourself that you never send. Anywhere the work cannot be judged, because judgement is what makes people stop early.

I keep a file called **scraps**. Nothing in it is organised. Nothing in it is meant to survive. Perhaps one line in twenty ends up in something real, and that ratio is fine, because the other nineteen were the cost of finding it.

## Then be ruthless

The permission to write badly comes with an obligation to edit honestly. Once the argument is there, the generosity has to point outward. Cut the sentence you are fondest of if it is doing no work. Delete the paragraph that only exists because it was hard to write. Read it aloud, and wherever you stumble, the reader will stumble too.

Two passes, two mindsets. The first one is for you. The second one is for everyone else. Most of the pain in making things comes from trying to do both at once.

The first draft is a conversation with yourself

Our studio has good tools. We could open any of them and start moving rectangles around within about four seconds. Instead, most weeks begin with a table, a stack of cheap paper and a box of pencils that nobody keeps track of.

It is not nostalgia. Paper does three things that a screen quietly refuses to do.

## It refuses to look finished

A sketch is obviously provisional. Nobody argues with the corner radius of a pencil line, and nobody assumes a decision has been made just because something appeared on a page. That looseness buys a room full of people permission to disagree early, while disagreeing is still cheap.

On a screen the opposite happens. Type is set, spacing is even, colours are chosen, and suddenly a rough idea is wearing a suit. People respond to the suit. They start editing details of a thing whose shape is still wrong.

## It keeps the whole problem visible

Twelve sketches can sit side by side on one table. Twelve screens cannot. There is no scrolling, no switching, no window on top of another window. Everyone in the room can see the full range of what has been considered, which is the only reliable way to notice the option nobody has drawn yet.

## It is fast enough to be honest

The gap between having a thought and seeing it is roughly one second on paper. That speed matters more than fidelity, because it means a bad idea can be tested and abandoned before anybody has invested enough effort to defend it.

- Draw six options, not one.
- Give each one ninety seconds.
- Say what is wrong with your own before anyone else does.
- Photograph the table, then throw the paper away.

## Where it stops

Paper is a terrible place to end. It does not scroll, resize, load slowly, or fail on a narrow screen, and every one of those things is where real design decisions live. So the sketch is a starting gun and nothing more. We usually spend one morning on paper and the rest of the project in the browser, which sounds like a small ratio until you count how many arguments that morning prevents.

Why we still sketch on paper

I have worked on teams of forty and teams of four. The teams of four were better, and for a long time I assumed the reason was talent. It was not. It was the number of conversations required before anything could happen.

Four people have six possible pairs. Forty people have seven hundred and eighty. Coordination is not a soft cost, it is the cost, and almost every process a growing company adds is an attempt to pay it without admitting it.

## The standing agreement

The best month of work I remember came from a rule we set in about ten minutes and then never broke: **anyone can decide anything reversible on their own.** If undoing the decision would take less than a day, you did not need a meeting, a document, or permission. You needed to tell everyone afterwards.

It sounds reckless. In practice it moved perhaps two decisions a week out of a meeting and into someone's afternoon, and the handful we got wrong were, by definition, cheap to undo.

## Meetings as a last resort

We kept two, and killed the rest.

1. Twenty minutes on Monday, standing, to say what each of us was doing and what was in the way.
2. An hour on Thursday, sitting, to look at real work on a real screen and be honest about it.

Everything else was written down. Written updates are slower to produce and dramatically faster to consume, and they have the useful side effect of exposing vague thinking, which a confident voice in a room can hide indefinitely.

## Protect the quiet

The part people underestimate is the silence. Making something good requires holding a lot of context in your head at once, and that state takes roughly half an hour to reach and about four seconds to destroy.

> A day with three meetings in it is not a day with five hours of work left. It is a day with three fragments.

So we defended mornings. No calls before noon, no expectation of an instant reply, and a shared understanding that a message sent at ten was being read at one. Nothing broke. Quite a lot got finished.

## It does not scale, and that is fine

I am not claiming this survives a company of two hundred. I am claiming that most teams are far smaller than the process they have adopted, and that shedding it is usually available to them the moment somebody is willing to ask.

Small teams, short meetings, long stretches of quiet

I want to be careful here, because slow software usually means the wrong thing. Nobody is arguing for a spinner. Speed of use is close to sacred, and a tool that wastes a second of someone's attention every day has taken a week of their life over a decade.

What I mean is slower to change. Software that is permitted to be finished.

## The cost of motion

Every release asks something of the person using it. Where did that button go. Why is this now three clicks. Which of my habits still works this morning. Individually these are trivial, and collectively they are the reason people describe perfectly capable tools as exhausting.

The teams making those releases rarely feel the cost, because they experience each change once, at the moment they build it. Users experience the accumulated total, all at once, every time they come back.

## What stillness buys

- **Depth.** People only develop real fluency with a tool that stops moving long enough to be learned.
- **Trust.** A thing that behaves the same way on Tuesday as it did in March becomes something you stop thinking about, which is the highest compliment a tool can be paid.
- **Better decisions.** When shipping is not automatically a virtue, the argument for a change has to stand on its own.

## How to slow down without stopping

The version of this I actually practise is not restraint for its own sake. It is a division: keep the surface still and let the inside move as much as it likes.

Performance work, reliability, security, the quiet removal of things nobody used. All of that can ship every day and nobody has to relearn a thing. Changes to the shape of the product, the vocabulary, the layout, the place where the important button lives, those get grouped, considered, and released rarely enough that a person can notice them.

> Ask of every visible change: is this worth the cost of being new to somebody?

Often the answer is yes. Frequently it is no, and the change was really about the team needing to feel motion. That is a real need, and it deserves an honest name rather than a release.

## The tools I love

Almost everything I use daily and would fight to keep has been essentially the same for years. A text editor. A camera. A note-taking app that has resisted every temptation to become a workspace. None of them are exciting. All of them are, at this point, load-bearing.

A case for slower software

There is a moment near the end of most projects when someone realises the empty state needs words. It is late, everyone is tired, and the words get written in four minutes by whoever has the file open. Those four minutes then represent the product to every person who ever hits that screen.

Copy is not a coat of paint. It is the only part of the thing that talks.

## Say the true thing first

Most weak interface writing is weak because it is postponing something. It softens, it apologises, it explains the mechanism instead of the consequence. The reader has to translate.

Compare a message that says an operation could not be completed with one that says the file is too large to upload, and that the limit is twenty megabytes. The second is not cleverer. It is simply the sentence someone was avoiding.

## Shorter, but not clipped

Brevity is a good instinct that curdles quickly. Cutting words is only worth doing while meaning survives, and past a certain point you are just handing the work back to the reader.

- Cut adverbs before you cut nouns.
- Keep the word that carries the specific detail, always.
- If a sentence needs a comma to breathe, give it one.
- Read it aloud. The stumble is the edit.

## Rhythm is a real thing

Sentences of the same length in a row flatten out and stop being heard. Vary them. A long one that carries an idea through a couple of turns, and then a short one. The short one lands.

> Everything you write has a sound, whether or not you chose it.

## Vocabulary is a promise

Pick a word for each concept and never quietly swap it. If it is a *project* on Monday it cannot be a *workspace* on Thursday. Every synonym asks the reader to check whether something new has been introduced, and that check costs more attention than the variety was worth.

Keep the list somewhere visible. Ten or fifteen words is usually the whole vocabulary of a product, and getting them stable does more for clarity than any amount of polish applied afterwards.

## Write it first

The best trick I know is to write the copy before the layout. Empty state, error state, the confirmation, the one line that explains what this is. If the words are hard to write, the idea is not finished, and it is far cheaper to learn that on a page than after the thing is built.

Notes on making things carefully.

The first draft is a conversation with yourself

More writing

Why we still sketch on paper

Small teams, short meetings, long stretches of quiet

A case for slower software

The shape of a good sentence

Notes on making things carefully.

Featured note

The first draft is a conversation with yourself

More writing

Why we still sketch on paper

Small teams, short meetings, long stretches of quiet

A case for slower software

The shape of a good sentence

Get the next note in your inbox