I went into Nicolas Darveau-Garneau's Be a Sequoia, Not a Bonsai ready to agree with it. The pitch lands fast if you run a small company: stop pruning yourself into a tidy little shape, grow into something that lasts. I had most of an essay in my head before I finished the introduction.
Then I checked the tree.
The bonsai is supposed to be the cautionary tale here, the thing that looks perfect and dies early for it. It doesn't. Bonsai routinely live as long as the same species planted in open ground, and often longer, because someone waters them every day and catches the disease in week one. The oldest tree in the US National Arboretum's collection is a Japanese white pine that has been in training since 1625. It was sitting in Masaru Yamaki's garden about two miles from the hypocenter when the bomb fell on Hiroshima. It survived. The museum didn't even learn that part until 2001, when his grandsons mentioned it during a visit.
So the metaphor is broken in the place where it was carrying the argument.
What the bonsai actually is
A bonsai is a maintained thing. Root-pruned on a schedule, wired, trimmed, repotted, held at precisely the size it already is. Four hundred years of skilled attention spent preventing growth.
That version is worth being afraid of, and it's much closer to how small companies actually stall out. Twenty years of competent work keeps the thing alive at exactly the size it started, because every hour goes into maintenance and none of it accumulates. Nothing dramatic happens, which is the problem. You can run a business that way for a very long time. Plenty of people do, and some of them are genuinely happy. I'm not going to pretend it's always the wrong call.
But it is a call, and most founders make it without noticing they made it.
The sequoia half holds up
Here's where the book is right, and the evidence is better than the metaphor it arrived in.
In 2014 a team led by Nate Stephenson published a study in Nature covering 403 tropical and temperate species and 673,046 individual trees, some with measurement records going back more than eighty years. For 97% of those species, the rate at which a tree puts on mass keeps increasing as the tree gets larger. Old trees don't slow down. A single big one can add as much wood in a year as an entire mid-sized tree contains.
Growth compounds with size. That's a measured fact about trees, and it's the whole argument. The bonsai imagery was decoration on top of it.
The split that matters
Founders hear "sustainable growth" and quietly translate it as stay small, stay safe, keep everything neat. I've done it. It feels like discipline while it's happening.
The useful distinction is between work that has to be repeated and work that stacks.
Pruning gets repeated. You do it, it grows back, you do it again in March. That's most of what fills a week: the invoice you re-chase, the report you rebuild from the same four sources, the reply you've now written a fourth time this month. Systems stack. Once the thing runs, it keeps running, and next year you build on top of it rather than rebuilding it.
This is most of what I do with AI now, and it's why I keep saying it belongs in the plumbing. A drafting loop that takes an hour off every Monday is worth more in year three than a chatbot that impressed the room in week one. The hour comes back every Monday. The applause doesn't.
Aggressive in method, patient in metric. Move fast on the experiment. Judge it on a number that takes a year to move.
The part I keep chewing on
Sequoias fall over. They get top-heavy, the root plate is shallow relative to all that mass, and a tree that stood for two thousand years is horizontal by the afternoon. The book doesn't dwell on it and I understand the choice, but I think about it more than I think about the pruning.
I don't have a tidy resolution. The bonsai keeper knows exactly what he's doing every spring, and that's worth something real. Most of the founders I meet don't know which of the two they picked.
The Yamaki pine has been in training since 1625. Someone will prune it again this spring.