The Seedbed — a commonplace book of unfinished ideas

For a while I kept two notebooks, and it made me stupid in a very specific way: every time I sat down to write, the first decision was not what to say but where to say it. A short thought felt unworthy of the long book; a long thought got trimmed to fit the short one. The taxonomy was eating the writing before the writing could happen.

So this page is one stream now, and the rule is that there is no rule about length. A single line is an entry. Four paragraphs with a footnote is an entry.¹ The kind is a description applied after the fact — a label on the jar, not a decision about which jar to buy. I have never once regretted writing something down; I have frequently regretted the sorting systems I built instead.

The quiet benefit is chronological honesty. When the small things and the large things share a spine, you can see what actually fed what: the line in June that became the argument in July. Sorted into separate ledgers, that lineage is invisible. Kept together, the page reads the way thinking actually goes — unevenly, with seeds and essays next to each other, and no apology for it.

Every few months I want to begin again from nothing: new notes, new system, new folder with a confident name. The urge always arrives dressed as clarity, and it is almost always avoidance.

The honest version of the impulse is that the old material has become hard to face. Not disorganised — faced. There are half-finished arguments in there, and opinions I have quietly outgrown, and evidence that I was wrong in writing. A clean slate makes all of that disappear without the work of disagreeing with myself.

What I do instead now is smaller and slower. I keep the old leaves and add a line beneath the ones I no longer believe. The record stops being a monument and becomes a conversation, which is harder to be proud of and much more useful to read.

A slate that is never wiped accumulates something a fresh one cannot: proof that I have changed. That is the only kind of progress I have learned to trust.

Time is a river which sweeps me along, but I am the river. ¹

I keep returning to this line. The river image only works because the water is moving — a still river is just a pond. The honesty of a clock is the same: it refuses to let the moment pool.

A good tool does two jobs. It finishes the work in front of you, and it changes what you believe the work is. The first job is the one we shop for. The second is the one that shapes us.

A hand plane teaches grain. A debugger teaches that a program is a sequence of states rather than a paragraph of intentions. Neither lesson appears in the manual, and both survive long after the particular task is forgotten.

So I have grown suspicious of tools that finish the work without teaching anything. They are not evil; they are simply expensive in a currency that does not show up on the invoice. When the tool leaves, the understanding leaves with it, and I am exactly as capable as I was before I began.

The test I apply now is small: after using this thing for a month, will I know something I could explain to another person? If the answer is no, I try to keep the tool at arm’s length, and do at least one pass by hand.

There is a shelf here of things abandoned: two half-built programs, a translation that stalled at the third page, and a set of notes for an essay I no longer agree with. For years I thought of it as evidence of failure.

I have changed my mind. The shelf is where the ideas go that were not wrong, only early. Twice now something from it has come back and finished itself quickly, because the difficult part had already been done and left to cool.

The condition is that the shelf must stay visible. Abandoned work hidden in an archive is simply lost; abandoned work in plain sight is a standing offer. I walk past it most days, and every so often it accepts.

The useful thing about a notebook is not that it remembers. It is that it disagrees. Memory is a diplomat; it smooths the record until the past appears to have led sensibly to the present. Paper has no such manners.

I went back this week to entries from four years ago, expecting continuity, and found a person confidently certain of two things I now think are false. There was no gradual revision in between. The change happened somewhere off the page, and I would have sworn under oath that I had always believed the later version.

That is the whole case for keeping a record: not nostalgia, and not productivity, but the ability to be contradicted by primary evidence. It is the cheapest available protection against becoming a person who has never been wrong.