The Maintenance Era

Maintenance is what longing looks like when it commits to time.

Spouse ride over Colorado in a Chinook - maintained to be safe enough for the lives of our soldiers.

This begins as a reclamation. Thirty years with my spouse Brent — built on luck and grit and timing we were smart enough not to waste. We’ve made it through deployments, wars, loss, plenty. Not because we’re exceptional but because we knew how to maintain. Tend. Stay present to our longing for a home and a life that was good and kind and generous.

When people hear “maintenance test pilot” they think of the aircraft. That’s part of it — Brent has tested not just whether planes fly but whether the protocols for keeping them flying will hold. He can predict what will happen because his routines are that dialed in. That’s the work that keeps four hundred people alive at 35,000 feet, and nobody applauds because the whole point is that nothing goes wrong.

But the maintenance isn’t just the aircraft. It’s us. We maintain our household through logistics — menus planned so there’s one store trip, budgets tracked, daily and monthly and quarterly checklists for the house. That’s the scaffolding. It frees us up for the actual thing, which is harder to put on a list: sitting still together, walking, exploring the never-ending curiosity of our minds. My husband and the way he thinks has thrilled me in ways I can’t fully explain for three decades. It’s the talking and the exploring and the embodied love of building something together and tending it. He taught me discipline in body, mind, and spirit. Every day I don’t want to do something but do it anyway, I’m rewarded in spades. Compound interest — not from the checklists, but from the presence the checklists made room for.

Everything I know about maintenance, I learned first in this partnership. The rest came later — boats, aircraft, software, the company I run. But it started here, in the daily practice of showing up to something you want to keep going.

My friend Dawn sent me a link in 2022. I took one look — boats, very long read — and closed the tab. If I want to read, I’ll open a book.

A few days later I asked for the link again. Dawn is wise, and I’d been turning it over despite myself. I sat in the sun on a Friday and read Stewart Brand’s “The Maintenance Race” on a site called Works in Progress. She was right. It’s well written.

I didn’t know then that the piece was a draft — published deliberately unfinished, open for comments, what Brand called a shakedown cruise. I didn’t know it would become the first chapter of his book Maintenance of Everything, which I’m reading now, three years later. I didn’t know I was picking up a thread I’d still be holding.

But that’s how maintenance works. Someone hands you something. You carry it. Sometimes you forget you’re carrying it until the weight shows up again and you realize it's shaped how you walk.

Brand tells the story of the 1968 Golden Globe race — nine men trying to sail solo around the world without stopping. No resupply, no help. Everyone remembers it as a contest of endurance. Brand argues it was a contest of maintenance styles.

Robin Knox-Johnston won. His boat was 32 feet of simple wood. He’d already sailed her 17,000 miles and knew every plank. He spent three hours a day keeping her sound enough to get home. He’d learned knots and splices and canvas work from a bosun in the Merchant Navy — the kind of apprenticeship that looks like drudgery until it saves your life. His thorough preparation equipped him to improvise.

Bernard Moitessier was the best maintainer in the race. He glued sextant legs with epoxy, replaced sail lashings, spliced halyards — and wrote about it the way some people write about gardens. His reward for a boat that functioned like new every day was time to read, sleep, and eat. When he reached the point where he could have claimed the trophy, he kept sailing toward the Pacific. He didn’t want to win. He wanted to keep tending.

Donald Crowhurst died. He rushed his preparation, launched in an unfit trimaran, and faked his position reports for months. Brand writes that optimists like Crowhurst tend to resent maintenance. Crowhurst called it “sailorizing.” He’d make task lists, do a few things half-heartedly, reward himself with a drink, and lose interest. When the distance between his reports and his reality became unbearable, he walked into the sea.

Three relationships to maintenance. Three outcomes.

I highlighted a line from that piece: “Skilled maintainers advise never trying to solve a new or complex problem without a thorough mulling first.” The man who said it was having a cigarette. He wasn’t in a hurry. He was being responsible to the problem before picking up a tool.

I highlighted another: “Doing maintenance cures depression.”

And I wrote, that summer, that the story commemorates the audacity of humans — the competitiveness, the incentive, the reward. That all of it was wrapped in a philosophy I was just learning: that doing things over and over again is what is truly exceptional.

I still believe that. But I’d push it further now.

One generation. That’s all it took to build a world that optimized for replacement over tending. Buy, use, discard. Ship, sell, deprecate. We trained ourselves to expect the new thing rather than maintain the current one, and we called it progress. The bill for that choice is coming due — in our infrastructure, in our institutions, in the systems we’ve asked to carry weight they were never maintained to bear.

And about those AI systems that “hallucinate” — let’s be precise about what happened. Teams built large language models using statistical brute force without sanitizing their inputs. They layered on weights to make the outputs sound conversational — warm, confident, human-adjacent — so convincing that people started talking to the software like it was thinking. Then, when the models produced confident-sounding garbage, the industry coined “hallucination.” Not “error.” Not “bad output from dirty data.” Hallucination — a word that implies the machine was almost right, almost conscious, just slightly off. The most successful piece of marketing the AI industry has ever produced. The machine isn’t hallucinating. It never was. It’s doing exactly what it was built to do: generate statistically probable text without any infrastructure to verify whether that text is true.

The fix isn’t better prompting or more parameters. It’s maintaining the knowledge underneath. Building systems where every claim has a traceable origin, where you can follow the lineage from output back to source and see whether the source still holds. That’s not an AI problem. It’s a maintenance problem. And it’s what we build at Bast AI — not a tool that generates answers, but an architecture that maintains the relationships between what is known, what has changed, and what was never true in the first place.

Maintenance is sometimes duty. Sometimes grit. Sometimes just the quiet labor of Wednesday morning. But underneath all of it there’s a longing. A wanting for the thing to remain viable, to keep going, to grow into something you couldn’t have built in a single burst.

We have this capacity as humans — to take raw desire and transmute it. To alchemize longing into practice, practice into resilience, resilience into something that compounds over years until it barely resembles what you started with but contains all of it. That’s what thirty years looks like. That’s what a maintained boat looks like after it rounds Cape Horn. That’s what knowledge looks like when someone tends the connections instead of letting them rot.

Maintenance is what longing looks like when it commits to time.

Dawn sent me a link. I forgot it was becoming a book. I carried the idea anyway — into a company, into a product, into an argument about what AI infrastructure should be. I didn’t maintain the citation. I maintained the thread.

That might be the most honest thing I can say about the maintenance era: it isn’t coming. It’s already here, in the things we’ve been tending without knowing it. The question is whether we’ll build systems that honor that work — or keep pretending the next new thing will save us from the last one.

It starts with someone handing you something worth tending.

(Author’s note - this is a remix of a piece I wrote in 2022 - Medium link for posterity)