gently illuminated by the memory of hope
On joy, profit as a value, woodworking, and the totally predictable consequence of speeding up a millstone.
In a vulnerable conversation with a friend recently, we both admitted that we're tired of working in tech. The tone was one of bemused resignation. What else is there to do at this point except sip cocktails in the quicksand together? Our species (tech professionals, I mean) doesn’t normally talk like this. Usually, we do the grown-up talk about tools and strategies and optimization (and lately, we talk ad nauseam about AI). We don't sit there and say out loud that we're tired in some fundamental way that a vacation can’t fix, and then 30 minutes later hop into a client call like nothing happened and this is a completely normal reality we’re all living in.
Every time I log into LinkedIn there are half a dozen new tools everyone is now saying are essential, plus several new workflows I should be adopting. Just today, Anthropic delivered new model Fable 5 unto the world. I read the news release, sat back in my chair, and just stared at my hands for a minute (and then, of course, I spent the next 2 hours testing it).
Forgive me for indulging in nostalgia… but it wasn't like this fifteen years ago :shakes-fist-at-sky. Breaking version updates would happen occasionally and cause a couple of weeks of chaos, and maybe you’d spend some time learning a new javascript framework that everyone was on about, but the newness had time to settle and be absorbed. For the past several months, I’ve felt like I'm running on something that's moving faster than I am, and I can no longer tell if I'm actually getting anywhere or just not falling down.
About 10 months ago, I think I subconsciously believed AI was going to fix my life. Not the industry (certainly not the economy), just my life, specifically. I had this picture in my head that I'd automate the tedious parts of my job, the fifty percent that's essentially translation work, taking someone else's specifications and converting them into code. I thought I would keep the architecture and creative problem-solving I actually care about and just offload the rote, mechanical part. I have languages and schematics in my head, and my job most days is to take these often unimaginative scopes and translate them directly into one of those languages.
And I imagined that once AI handled the translation, I'd have all this time. I’d be able to embrace more interesting projects, and I’d spend more time in the artistic and relational parts of my work. I'd go on a walk after lunch while a machine wrote my code for me. I'd mentally clock-in about four hours a day, and the rest would just be mine.
That fantasy felt so real to me that I built my entire emotional relationship to AI on it, on the promise of relief. I was drowning in work like a skier overtaken by an avalanche, and AI looked like the thing that would finally let me come up for air.
….
AI has saved time. My team shrank (no layoffs, just natural movement), our client roster increased, and magically we’re handling the additional workload without needing to hire new staff. Three years ago, we would have added 2 more full-time engineers by now.
Except the time I saved doesn't belong to me.
It belongs to whoever sets the pace, and the pace is not set by the clock or by the work. It's set by market pressure and the ambient understanding that if you can produce more, you should (and fast), because if you don't, someone else will. The efficiency I created didn't come back to me as leisure or breathing room. It came back as expected throughput. I'm the pipe the gains flow through, not the beneficiary to whom they are delivered. My review queue now feels like 10 developers' worth of output, because that's essentially what it is, and whatever hours AI freed up, the general conditions of work in consumer tech reclaimed them before I even appreciated they were gone.
I sped up the millstone. The mill stayed exactly the same.
Though I'm tired down to my bones lately… I really like my job. I adore my colleagues, and I relish the opportunities for craft, service, achievement, and relationship. As tech careers go, I have a good thing going. I have significant autonomy, everyone I work with is awesome, our clients are lovely, and every day is different.
On a fundamental level, I have always derived profound satisfaction from writing complex logic that actually works. The first time I wrote a program that made a computer do something cool, I felt like a wizard. It was instantly addictive. That feeling is what pulled me into this career, and twenty years later it still fires the furnace.
But that feeling lives in one experiential layer, and the exhaustion lives in a different one. I lived for months around a growing, subterranean disquiet before it occurred to me to try to pry the two layers apart so I can understand what the hell is happening to me.
...
I realize now the work itself was never the joyless part. The joyless part is the conditions the work gets done under.
"Obviously," you are muttering to yourself. But I don't think I could see it as clearly until AI entered the picture and changed the work so dramatically.
Deadlines that compress everything into the minimum viable whatever. The pace: like if you stop moving, you die, and also if you stop growing, you die, and also you must both move and scale as fast as your mutant-shark body can go, and sleep when you're dead, and remark on how little of your "unlimited" time off you will be "able" to take this year. The painful disconnect between the elegance I have spent years learning to build into my product, and the total irrelevance of that quality to its market value. The mental downtime and creative flow time I desperately need and can never seem to argue for effectively enough.
I feel like a donkey in that flour mill. Eventually, after a few years hitched to the engineering wheel, most of us give up the awkward ambivalence and choose either resilience "I love my job, I’m going for staff/management" or escape “I’m rage-quitting and becoming a shepherd” because staying the nuanced and more genuine middle road is too confusing and requires admitting something that has no clear next step.
The usual story of technological progress is that automation comes along and kills the joy. It replaces human craft with machine output, and the people who found meaning in the doing are left managing processes they used to own, and they are sad, and everything gets standardized, less expensive, more available, and eventually shittier.
I think the joy was already mostly gone, the standardization was already embedded, and the creeping enshittification has been crawling over everything tech produces at scale for more than a decade.
Twenty years of this industry's worst impulses metastasizing - consuming human and natural resources and crapping out profit by the bucketful, over-leveraging venture capital into growth-at-all-costs - and the relentless optimization of everything, including the people doing the work - ground the joy out. By the time AI showed up, what it found was a workforce that was exhausted and overextended and, if we're being honest, ready to hand over whatever it could, because the work had stopped being something you enjoyed and had become something you survived. Of course you'd want to automate that. If your job already feels like working a factory line with algebra, then automation isn't a threat, it's a relief.
But AI can only touch the work. The conditions - the pace, the pressure, the inadequate rest, the “community” that seems composed of “hustle” and marketing bots instead of real human beings with hearts and indigestion and questions - all of that is social and economic. It's set by our wounded culture and hyperbolic markets and by the hidden characteristics of this bleak and dehumanizing paradigm we struggle to articulate but nonetheless continue to (apparently) collectively endorse.
AI has no access to that layer. It can only act on the other one.
When I reached for AI hoping for relief from the conditions, the only thing it could actually eat was the part I still loved. And whatever time it freed, the conditions swallowed whole, because it turns out it’s the conditions that decide how much gets demanded of me, not the difficulty or complexity of the work itself.
The tool I adopted to escape the joyless parts of my job could only consume the joyful parts.
The joyless parts of a job tend to be the parts you can write down completely; a spec with no creativity or unknowns left in it. Joy seems to live in the parts that resist being written down, the things you feel your way through rather than execute according to instructions. “Joyless” and "automatable” may be nearly the same property, seen from different dimensions.
We don't automate the joy away. I think we automate what is left after the joy has already been extracted.
I've been building electric guitars by hand for 10 years, and more recently I've gotten into traditional woodworking (hand planes, chisels, back saws, etc.) Traditional woodworkers use human-powered tools by choice. A CNC router can cut a dovetail joint that's technically flawless, and they simply don’t care. The process that produces the outcome matters as much as the outcome itself.
In actuality, the process is the fabric of which the outcome is composed (something we often either forget or pretend isn't true).
Power tools didn't kill traditional woodworking. They moved it from necessity to choice. What happened after that is telling: the craft lives on, not in spite of the machines, but almost because of them.
We know what the machines can do now, and what it’s like to use them, and because no one needs a coffee table yesterday, hobbyists can choose the quiet, a floor full of wood shavings, a sore back and hands, and the sensuality of elemental tools we push ourselves, over the roar, dust, danger, expense, endless setup, and byzantine safety protocols of a shop dominated by machines. Joy is remarkably resistant to automation - not because the machine can't do the work, but because the people who love the process don't want the machine to do it, and if people don't adopt the automation, it doesn’t matter how capable the machine (or the model) is.
I can see a day when writing code by hand becomes something like that. I suspect that when everything is built by machines, we’ll realize our software products all look the same the way mass-produced furniture all looks the same. It's already happening, even without AI. The effects of standardization can be functionally positive, but over time, we forget that our surroundings used to feel more inspiring.
The things we make carry the traces of us, the evidence that someone was there, thinking and choosing. Creative intent is always intertwined with anticipation; it is inherently optimistic. When we infuse ourselves in the process, the things we make remain gently illuminated by the memory of hope we leave behind.
Which brings me here, to this moment, where I am forced to acknowledge that the thing I have been putting the most of myself into for the last 10 months is, actually, artificial intelligence. In fact, the most interesting challenges of my entire career to this point are all mixed up in this technological marvel.
Building custom agents, pushing the limits of what autonomous systems can do, orchestrating coordination between programs that have never been connected before... it's a new, irresistible frontier. It feels like Manifest Destiny all over again, with potentially the same terrible result and the same imminent hangover, but my god, the allure of mapping a new wilderness is compelling. For the first time in years I feel like a builder and a visionary again instead of a resource slowly being extracted. I can't leave it alone. The work is so fascinating and so alien that I’m thinking about it in the shower, folding laundry, when I’m driving. I'm literally doing the thing I've spent the last 2000 words worrying about, and I cannot stop, and the reason I can't stop is because, all of a sudden, engineering feels brand-new. I’ve landed on the moon, and I can’t just go to bed at a reasonable hour like I’m still an Earthling.
I'm both afraid of and morbidly curious about the outcome of all this (e.g. are we accelerating the Star-Trek future, or the Mad-Max one?) But the temptation is wicked. Me and my laptop screen full of terminal windows can build things now that before I only dreamed of building in 2 years with a team of 10+ at a hypothetical company that would never hire me anyway. The frontier has given me back (temporarily) the parts the conditions took.
And yet, the fear that I have become the architect of my own obsolescence, that we are doing irreparable harm to the planet and to each other, and that my work is feeding the very profit-sovereign value system that is driving the breaking of everything that I value most, exists right alongside my excitement.
The most joyful work I've done in years is the work I'm most worried will ruin everything, and I’m just wandering around in the cognitive dissonance.
Come back next week to learn how I wired up the unauthorized MCP server I built for our project management software to a headless instance of Claude running on a dedicated Linux Macbook, so that our PM can now assign tickets directly to an agent, and I can sip my morning coffee while merging a week’s worth of non-human-authored PRs.
I’ll tell you all about how I downloaded my brain to a /senior-engineer skill and crafted the final-boss of meta-prompts, “be me, make no mistakes, ask clarifying questions before you begin.”
By August, I’ll be a shepherd.
