Why craft-lovers are losing their craft
17 points by hongminhee
17 points by hongminhee
If the grief Lawson describes is real (and I think it is), and if its deeper cause lies in the social relations around the technology rather than the technology itself, then the right target for that grief isn't the LLM coding assistant. It's whatever forces people to use tools they don't want to use, on terms they didn't choose.
This feels like a fallacy of assuming these are mutually exclusive. You can and should blame both!
Fair point, but I think the argument is less about blame and more about where change is actually possible. You can blame the hammer too, but the hammer doesn't change.
The reason I focus on the social relations is my own experience: same tools, different conditions (no employer, public funding), and the alienation just… isn't there. If the technology itself were the problem, that shouldn't work. So it seems like the more tractable target.
All of this is 100% reasonable, but how can you ignore the ethical, moral, legal, social, and psychological concerns other than alienation?
For me, the entire business model of the corporations releasing these "assistants" would be reason enough to boycott them even if I wanted or liked having such "assistance".
Those concerns are real, and I don't want to wave them away. But I'd push back on one premise: that the current exploitative development model is the only possible one.
LLMs trained on publicly funded data, with open weights, by a cooperative or public institution rather than competing corporations, would still be LLMs. The technology doesn't require the business model. That's actually the argument I've been trying to make in a piece from earlier this year.
Boycotting current vendors makes sense on those grounds. I do think, though, that “boycott the technology” and “change who controls the technology” are different strategies with different long-term implications.
It is not the only possible one but it is the one we have right now. How do we change who controls the technology without boycotting the people who control it?
Fair, and I don't think boycott is useless, just incomplete on its own.
A few things that could work alongside or instead: legislative requirements that publicly funded research produce open-weight models, public procurement rules that push governments toward open alternatives, compute infrastructure treated as a public utility (the “CERN for AI” argument), or data cooperatives that give people collective bargaining power over training data.
The licensing route I've written about before (“TGPL”) is another angle. Not a silver bullet, but the point is there are multiple levers, and they work better together than any one of them alone.
I apologize if I missed it but I don't think your post addresses the fact that AI companies assert that copyright-based licensing simply doesn't apply to them at all. They consider their activities to be permitted under fair use, which means they can use GPL, MIT, and "TGPL" licenced software alike without recompense or attribution.
It gives me a bitter chuckle to see this post - which is largely right, even if I don't agree (or even agree that Marx asserted) that smashing the looms was wrong - well-regarded on the same site where, right now, I'm being told that capitalism is the only viable economic system unless we come up with matter replicators.
If your plan to save the industry is to develop class consciousness among software engineers, you've already lost. Too many of us are permanently stuck in the mindset of the temporarily embarrassed millionaire.
That mindset exists, sure, mostly in the circles where Lobsters skews. I'm writing from Seoul, and it's much less dominant here than the comments section might suggest. Same goes for most of Europe, Latin America, and honestly most of the world where software engineers exist.
Treating “software engineers” as a monolithic class with uniform false consciousness is itself a bit Silicon Valley-centric. A lot of us are just workers. Not angling for a startup exit, not dreaming of stock options, just people who got into this because it was interesting and now need it to pay rent.
“Too many” having that mindset is different from “all of us do,” and the latter is what the argument would need.
I very much hope you're right. I currently work somewhere that is about as non-Silicon Valley as you can get in the US, and I still see the same problems with false consciousness - but I'd rather live in a world where programmers woke up to the reality of the situation, for sure.
Didn't the craft coders make other jobs redundant over the last 60 or so years?
I have vivid memories as a computer-loving kid having a family friend bitterly curse-out my fascination with these new devices -- because he was an old school architect and his inability to adapt to CAD had curtailed his career.
I want to be as compassionate as possible. I miss the craft as well, and there is grief there. These are real people and shift is jarringly rapid.
But I feel there is a real irony here that's being missed?
This one stings, and I think it's right.
Software engineers have spent decades building the infrastructure of automation. We told ourselves it was progress, and in some ways it was, but it also displaced a lot of people who didn't get to write think-pieces about craft and grief. The architect who cursed out your childhood fascination with computers had a point too.
Maybe there's something useful in this discomfort, though. If enough of us actually feel what it's like to have your livelihood threatened by a tool you didn't choose on terms you didn't set, that could build more solidarity with everyone else who's been through it already. Or we could just learn nothing and go back to not thinking about it once the immediate threat passes. History suggests the latter is more likely, but I'd rather not assume it.
Off-topic: curious how you're handling writing in multiple languages? You're fluent and just translated by yourself? Or did you use machine translation (seems like the Japanese one is potentially LLM-translated from the Korean one?).
First off, I write everything in my native language, Korean. Then, I run it through a machine translator to get a rough draft. My go-to translator these days is Kagi Translate. After that, I review the translation and edit it until I'm happy with it. Next, I use an LLM to check my edits for any grammatical errors or awkward phrasing that might sound unnatural to a native speaker.
This workflow works pretty well for me because my English and Japanese skills are asymmetrical when it comes to reading and writing. I'm quite good at reading, but I struggle with writing.
I continue to consider myself a craft coder even with the use of claude code. I can often roughly visualise up front the code that I want to produce, and either way I do examine every line that it generates, suggest (or sometimes manually perform) refactorings, etc, until I am happy with the code itself. it still feels like sculpting code into shape, versus targeting some output and not caring about how it is accomplished.