The Only Sane Place to Stand

The man with more identity invested in handwritten code than anyone alive just watched machines learn the craft — and he sounds grateful. This episode takes his stance apart: what actually becomes scarce when execution gets cheap, where attention lives while the future accelerates, how to let an old version of yourself die without calling it a waste, and why two teenagers in a kitchen are a better clock than any dashboard. Runtime about 40 minutes.

On a Sunday morning at the end of August I pulled the code repository behind this podcast and found a branch I had never seen. In the night, an agent had drafted the beginnings of an entirely different episode — research, an editorial brief, a pilot script. Nobody typed any of it. It was not perfect. It was Tuesday-good: work I would once have blocked out days for, done between midnight and five by something that does not sleep.

I would love to tell you my first feeling was wonder. It was a tightening under the sternum, and a question every builder I know is asking this year, mostly silently: am I behind?

That same week I listened to five and a quarter hours of David Heinemeier Hansson with Lex Fridman. David — programmers know him by his initials, DHH — built Ruby on Rails, the framework under Shopify and the early GitHub, and spent more than two decades as the industry’s loudest preacher of code as a handwritten craft. The machines have now come for exactly that. If anyone earned the right to sound bitter, it is him.

He calls this moment “the show of a lifetime.” He says: “You are alive in this moment where decades are happening in weeks.”

Either the man with the most to lose has misunderstood his own situation, or he understands something the rest of us keep missing. The episode is my attempt to take his stance apart and see what a finite person can actually use.

Cost was doing quality control

The economics first. When the price of an input collapses, value moves to its complements. The collapsing price is implementation — intention into working artifact. What appreciates: knowing what is worth building, formulating the real problem, taste, saying no, and responsibility, which no machine can carry, because responsibility is a promise with a person attached who can be wrong.

The honest numbers cut both ways. In a randomized study in spring 2025, sixteen experienced open-source developers were 19 percent slower with early AI tools — while estimating afterward that they had been about 20 percent faster. The feeling of speed is not a measurement. A year later the same lab could barely run the experiment: developers refuse tasks where AI is banned, and parallel agents defeat the stopwatch. Nobody can tell you where the line is. Which is why “am I behind?” has no answer — there is no longer a line to be behind.

But something real did change. It used to cost so much to build anything that most bad ideas died unbuilt. Cost was doing quality control. That filter is gone: a bad idea can have a working prototype and five polished variants before lunch, and the only thing left between an impulse and its artifact is somebody’s judgment.

When execution becomes abundant, judgment becomes expensive.

Where attention lives

I know the founder’s permanent future tense from the inside. The round lives in the future, the roadmap lives in the future, and the sentence that rules them all — when things calm down — I have said in January and in June. Planning is a tool. The failure is residence: when the future stops being a place you visit to make decisions and becomes the place you live.

In a classic experience-sampling study, minds were somewhere else in just under half of waking samples — and what people were thinking predicted their happiness better than what they were doing. A life can be enviable on paper and barely lived in.

I recognize the frontier side of this from the inside. I have agents running most days; my weekends are often an ocean of creativity, one thing flowing into the next — and that did not arrive with AI. I was like this with the 3D printers and the home automation. The machines did not create the appetite; they removed the ceiling. My one real superpower is deep focus: I can lose time and room completely, jump between wildly different things, tussock to tussock, and the line through them still points forward. I love it. But a superpower sends a bill, and mine gets sent to the three people actually in the room. Two wonderful children, a wonderful wife — meaning of exactly the same rank as the work, and the kind that only pays out in presence. I am not good at this yet. That is the plainest sentence in this episode.

Here is the paradox I keep turning over: the most future-facing builder in software refuses to extrapolate. David says that mental mode trends toward what he calls AI psychosis, and builds at today’s frontier instead. All the building happens at the edge of what actually exists. Nearly all the dread happens inside a forecast.

And the sharpest crack, which he shows himself: the most euphoric voice in the conversation admits the recent months were his most exhausting in five years, and that the pace cannot last. He sprints because he can see the end of the tunnel. A sprint is a strategy only as long as it has an end. The machines removed the natural stop — nothing makes you wait anymore — so the question quietly changed from how much can I get done to what deserves the time that just became available. My own best answer is small and daily: up at five, the run, the page, a book, no screens before six — and then the strangely productive hour between half past six and half past seven, the first screen hour of the day — an hour with a hard end, a sprint with an end, every morning, before the day has even started.

Letting a version of yourself die

Every morning after the run I write a diary, by hand, with my Lamy fountain pen in my Leuchtturm notebook. In a period of the company’s life when very little was working, I wrote a strange question at the top of one morning’s page: knowing everything, would I choose this life again, every wrong turn included? The answer came fast, and it was yes — not because it was pleasant, but because every discarded version of me had carried me to that page. There is an old name for that test; it is Nietzsche’s.

My own road here was never straight. Five years of a PhD that sounds tidy in a CV line and was anything but. Before that the navy, as a sea officer — many ships, many crews, many technical systems, and a large organization that teaches you the plan is not the journey. After the dissertation, the university years as what they call an intrapreneur: projects, collaborations, conferences, research funding, hires, a lab. Then a company. Then this one, and inside it, strategies we tried, believed in, and buried. The destination never moved; the routes kept changing — and every meter of that crooked road was walked in some present moment, by some version of me who could not see around the bend.

David takes the same test in public when Lex names what the agents mean for a life of mastery. No defense, no bitterness: gratitude, and amor fati. I think he can do it because he has separated three things we usually weld together. What the market pays for a skill. What practicing it made of you. And where it carried you. The machine repriced the first — brutally, maybe permanently. It cannot reach the other two. A craft can lose its market and keep its graduates.

To be fair about the cushion: David owns his company, and I own a piece of mine. It is easy to love your fate when your fate is compounding. What travels without the cushion is not serenity — it is the moves. Changing when the object changes. Separating what a craft paid from what it built: the translator whose rates collapsed this year still owns two languages and the judgment between them, and that is not consolation, that is capital. Refusing to move your residence into the forecast. You can copy the attention and the willingness to change. You cannot copy the bank account.

The versions end

David keeps one instrument he did not throw away: hidden behind the clock in his system, a quiet indicator of how much of a life is already gone. Memento mori as a background process. Whether that is morbid depends on which operation it performs: hurry spins the machine faster; what deserves my finite time re-sorts the goals. The psychology of shrinking horizons is remarkably consistent — people with visibly short horizons do not speed up. They get selective.

My children are fifteen and thirteen. And my son, right now, is deep inside one of the most joyful things I know exists: building his own AI from scratch, to understand it — writing the code by hand and running the machines flat out beside him, curiosity all the way down. The craft David spent two decades mastering and released with gratitude, Gustav is discovering fresh, in the same year, under the same roof as me, with the machines as playmates instead of rivals.

A calendar says the years passed; a child is the years, walking around the kitchen. My phone is full of photographic evidence of people who no longer exist — the four-year-old on the dock, the seven-year-old with the bicycle. Finitude is not only that my life ends. Every version of the people I love ends, on its own schedule, usually unannounced. Presence cannot be saved for later the way money can.

So, as plainly as I can say it: the company is not a waiting room for a success that makes the years retroactively worth it. The training is not a waiting room for a future body. The children are not a task to manage while the real life gets built. And this strange, accelerating year is not the waiting room where we sit until the real future gets announced. There is no announcement coming. The future keeps becoming the present while we prepare for it.

The future is arriving so quickly that the only sane place to stand is the present. Not because the future does not matter — because the present is where the judgment happens. I was born before the internet, and I am building after it. I was born before AI, and I am building with it.

What incredible luck, to be standing here while it happens.

Sources and footnotes: the full DHH–Lex Fridman transcript; the METR 2025 RCT and 2026 update; the customer-support field study (5,172 agents, +15 percent, largest gains for the least experienced); US BLS projections 2025–2035 (programmers −7 percent, developers +10 percent — forecasts, not fate); Killingsworth & Gilbert 2010; Rudd, Vohs & Aaker 2012 on awe and time (later replications mixed); Baron et al. 2017 on orthosomnia (three clinical cases); Carstensen’s socioemotional selectivity theory. This episode was made with AI throughout — script written with an AI collaborator, voice synthesized from my cloned voice, music generated. Given the subject, that is either irony or proof.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit frahlg.substack.com