I love the references to Jacques Ellul's ideas, which I think are interesting to reflect on in an AI age. It helps make clear that what is fundamentally at stake in much technological "progress" is an (often only tacitly acknowledged) sublimation of "efficiency" to the place of highest value.

What's fascinating is that this value elevation seems to have gone largely unchallenged, despite being in essence an arbitrary value choice. Other choices are possible. And, I hope, still are possible, despite what the bigcorps declare must be the case in order to maximize shareholder returns.

> But there are serious limits. [Your coding agent] will lie to you, they don't really understand things, and they often generate bad code.

I think that really high quality code can be created via coding agents. Not in one prompt, but instead an orchestration of planning, implementing, validating, and reviewing.

Its still engineering work. The code still matters. Its just a different tool to write the code.

I'd compare the difference between manually coding and operating a coding agent to the difference between a handsaw and a chainsaw - the end result is the same but the method is very different.

Enterprise software tends to particularly bad because it's being sold to managers who won't use it themselves. Consumer software tends to be more user-friendly (or it won't sell), but popular software isn't always what you want.

When writing software for yourself, there is a bias towards implementing just the features you want and never mind the rest. Sometimes the result can be pretty sloppy, but it works.

However, code health is a choice. You just need to know what to ask for. A coding agent can be used as a power washer to tidy up a project. This won't result in great art, but like raking leaves or cleaning your steps or plowing a driveway, it can be satisfying.

Just as you wouldn't use a power washer to clean a painting, maybe there's some code that's too delicate to use a coding agent on? But for a project that has good tests and isn't that delicate, which I believe includes most web apps, nobody's going to want to pay for you to do it by hand anymore. It would be like paying someone to clear the snow in a parking lot with a shovel rather than hiring someone with a plow.

Most software engineers were not “crafting” before AI. They were writing sloppy code for the sake of profit, getting a pay check, and going home. Which is why AI also outputs the same crappy code.

Rumor has it there were a few elite crafters among the lot. Software wizards who pondered about systems and architecture as they had a $10 espresso macchiato.

The AI code takeover will not free engineers up to do craftsmanship. It will annihilate the last vestiges of craftsmanship forever.

+1 for the mention of Forth. I use it often. LLM answers are possible now, but they are like translated C. It’s very bad style.

The standard: Forth words should be a few lines of code with a straightforward stack effect. Top level of a program might be 5 words.

LLM will generate some subroutines and one big word of 20-50 lines of nested IF..THEN..ELSE and DO..WHILE just as if it writing C.

I am in the middle of reading a fascinating book about the early days of computing: Turing’s Cathedral by George Dyson. It really opened my eyes to American engineering craft post WWW II.

We seem to take everything for granted now and forget what real engineering is like.

This review is 13 years old by itself:

https://www.theguardian.com/books/2012/mar/25/turings-cathed...

I feel like one of the things that's not said enough, and which I think is conflating the effectiveness of AI in the eyes of actual software engineers, is that, for the most part, most code produced IS lousy. The craft of programming has been watered down so much in favor of results, and so much code is disposable or write-and-use once, that quality just became less relevant.

I remember when I first started out programming 20 years ago, there was time to craft good quality code. Then there were more and more pushes to get more code out faster, and no one really cared about the quality. Bugs became part of the cost of doing business. I think GenAI for code fits well in the more recent paradigm, and comparing it with hand-crafted code of yore is a bit disingenuous, as appealing as it may be, because most code hasn't been that good for a long time.

I am sad to admit it, but AI is just fitting in where poor coding practices have already existed, and flourishing in that local maxima.

This really nails the core issue: AI thrives in environments where software is treated as “good enough” optimization rather than craft. It’s not replacing great engineers so much as exposing how much of modern software has already become rote, metric-driven work. The Arts & Crafts parallel feels especially apt as mass-produced code gets cheaper, human judgment and taste become the real scarce resources.

I think we should embrace AI to craft better software. You have a lot of control over the code generated by AI, so all your designs, patterns, best practices can be used in the generated code. This will make us better software craftsmen.

A nice example is guitar building: there's a whole bunch of luthiers that stick to traditional methods to build guitars, or even just limit themselves to japanese woodworking tools.

But that is not the only way to build great guitars. It can be done by excellent luthiers, building high quality quitars with state of the art tools. For example Ulrich Teuffel who uses all sorts of high tech like CAD systems and CDC machines to craft beautiful guitars: https://www.youtube.com/watch?v=GLZOxwmcFVo and https://www.youtube.com/watch?v=GLZOxwmcFVo

Unfortunately, craftsmanship does not come cheap, so most customers will turn to industrially created products. Same for software.

Great article, but doesn't address the fundamental issue: defining quality. Other than some objective metrics like code coverage, there is little agreement about what constitutes good code. The closest thing to a consensus might be the rules encoded in linters/formatters. Each Rubocop or eslint rule had to go through code review and public scrutiny to be included and maintained. Most often the rules are customized per project/team. Of course this runs into the same problem the article mentions: narrowness of vision. It seems the only way to achieve a high-minded ideal is the BDFL model of software development.

AI takes jobs faster than creating new ones.

That's the problem.

Stop arguing about whether AI ruins craft. Fix the incentive structures so that increased productivity buys humans time, not disposability. Then see what kind of craft emerGe

I don't get why the author assumes AI-assisted coding can never produce elegant software. I am learning new things from AI almost in every interaction.

"But attempts to generalize their capabilities have largely failed and produced code that is novel and impressive only in its monstrosity[link gastown]" was very funny thanks

This argument is basically just the 1800s Luddite vs Industrialist argument recast for a new age. Group A thinks quality is about human agency, and that machines are being used to bypass the apprenticeship system and produce inferior goods. Group B thinks efficiency is the highest priority, and craft is just vanity. Of course as we know we went a third way, and human roles just shifted.

I think one promising shift direction is humans do NOT like to talk to bots, especially not for anything important. It's biological. We evolved to learn from and interact with other humans, preferably the same group over a long time, so we really get to understand/mirror/like/support each other.

On a side track, I wish to express my fears regarding AI

Unfortunately for the general populace, most technological improvements in information technology, for the past 5 decades, has lead to loss of political control and lessened their leverage for political change.

With AI, this change is going to be accelerated a 100 times.

With current AI slop, and more importantly, almost indistinguishable from reality, AI based content, the populace is going slowly learning to reject what they see and what they hear from mass media.

AI has muddied the pool so much, that every fish, us, cannot see the whole pool. What this will lead to, is for political figures and bad actors to, much more easily almost with no effort at all, create isolation among people.

No event will create a mass uprising, because no event can be believed by a common mass. It will be easy to generate an alternative reality using the same AI.

Now, the political class and the billionaire class, are free to act with impunity, because the last check on their power, the power of mass media to form public opinion, to inspire the masses to demand change or accountability, has eroded to the point of no return. (They have already captured the institutions of public power)

I fear for the future of humanity.

Edit : There are already troubling signs from the billionaire class regarding this. There is a narrative to "ensure guardrails" for AI, sort of giving the populace the idea that once that is done, AI is acceptable. This is like saying, "better have a sleeve on the knife, so that no one can cut with it, but use it as a prop in a movie"

They are creating this narrative that AI is inevitable.

They are fear mongering that AI is going to take jobs, which it will, but it also goads the capable ones to get on to the bandwagon and advance AI further.

Agreed. AI-generated code is "mid" by nature. You won’t feel amazed seeing AI-generated code because its "reasoning" competes with that of a potato.

I have recently started exploring AI-coding -- note that I said AI coding and not vibe-coding because that is for the brain-dead.

By AI coding, I mean I know the inputs, outputs, structures the code should have and the necessary context to write the code. Then, articulating the requirements in English as best as I can and feeding it to agents.

Needless to say, the code is pathetic; it chooses to implement meaningless abstractions even after explicitly providing the design and plan to follow.

I don’t understand how we, as a collective species, agreed to believe the criminally wrong lies of tech CEOs that, instead of implementing a “reliable system by hand,” we choose to convey our ideas and instructions in an “ambiguous,” “inconsistent,” and “context-dependent” language (English), which is then passed through a probabilistic system to generate the reliable system.

AI wrote better code than most of my colleagues.

Especially with my rules:

- Prefer simple, boring solution

- Before adding complexity to work around a constraint, ask if the constraint needs to exist.

- Remember: The best code is often the code you don't write.

> But there are serious limits. [Your coding agent] will lie to you, they don't really understand things, and they often generate bad code.

As for lies and bad code, it didn't appear with AI. Humans lied and produced bad code before AI.

How does the author empirically know AI does not understand? And if it does not understand right now, is a machine fundamentally unable to understand? Is understanding an exclusive human ability? Is it because machines lack a soul? It sounds quite dualistic (Descartes'view that mind and body and fundamentally different).

Don't get me know, I think right now, AI is less a good at understanding humans than other humans (or even dogs) in many contexts because it has no access to non verbal signals. But in the context of building software, it is good enough and I don't see why a machine should not be able to understand humans.

What if AI starts to have sense of craft? we just miss the verify and critique models, that will tell other models what looks good

Many advances have their slop version, look at any trend which began as broadly positive and became infested with a tacky parade of bandwagon pretenders.

AI is just the next example, and the internet is particularly troublesome at filtering up the most tacky.

Always remember the broadly positive aspects remain, they're simply obscured by dusty clouds from the bandwagon.

Use tools to expand your capability.

> People have said that software engineering at large tech companies resembles "plumbing"

> AI code [..] may also free up a space for engineers seeking to restore a genuine sense of craft and creative expression

This resonates with me, as someone who joined the industry circa 2013, and discovered that most of the big tech jobs were essentially glorified plumbers.

In the 2000s, the web felt more fun, more unique, more unhinged. Websites were simple, and Flash was rampant, but it felt like the ratio of creators to consumers was higher than now.

With Claude Code/Codex, I've built a bunch of things that usually would die at a domain name purchase or init commit. Now I actually have the bandwidth to ship them!

This ease of dev also means we'll see an explosion in slopware, which we're already starting to see with App Store submissions up 60% over the last year[0].

My hope is that, with the increase of slop, we'll also see an increase in craft. Even if the proportion drops, the scale should make up for it.

We sit in prefab homes, cherishing the cathedrals of yesteryear, often forgetting that we've built skyscrapers the ancient architects could never dream of.

More software is good. Computers finally work the way we always expected them to!

[0]https://www.a16z.news/p/charts-of-the-week-the-almighty-cons...

The framing of craft vs. slop misses something important: most production software quality problems aren't about aesthetics or elegance, they're about correctness under real-world conditions.

I've been using AI coding tools heavily for the past year. They're genuinely useful for the "plumbing" - glue code, boilerplate, test scaffolding. But where they consistently fail is reasoning about system-level concerns: authorization boundaries, failure modes, state consistency across services.

The article mentions AI works best on "well-defined prompts for already often-solved problems." This is accurate. The challenge is that in production, the hard problems are rarely well-defined - they emerge from the interaction between your code and reality: rate limits you didn't anticipate, edge cases in user behavior, security assumptions that don't hold.

Craft isn't about writing beautiful code. It's about having developed judgment for which corners you can't cut - something that comes from having been burned by the consequences.

Enterprise software is also particularly bad because many of the customers get to demand that things work the way they want. Leading to a million weird functions, toggles, configurability because some manager in charge of making a big purchase demanded that first it must do X, many of these features left with not even a single user after the original requester leaves. While consumer software the individual consumers just get what they are given, and a single product manager/team decide what's best.

* Enterprise software tends to particularly bad because it's being sold to managers who won't use it themselves.*.

Don’t forget that managers have different goals than file and rank employees.

For SaaS I work for we get requirements like required fields for a process that manager needs to have correct data and for insights into business process.

After we deliver software we get support tickets from employees that are using system nagging that “it takes too much time to fill in all this data” and that we should “fix our shitty system”.

They don’t care and they don’t have full knowledge why stuff is required - which is fault of managers that are not training their people and explaining “why”.

Oh and of course they have to copy paste shit over and over because their company won’t have budget for us integrating with their CRM and we won’t invest in something that benefits only single customer who might not renew the license next year - but also they don’t want to make a commitment like 5 years contract where we could do some investment. Of course there are some that invest in connecting the CRM but it mostly is an exception rather than the rule.

Consumer software can be good, but it's often also optimized for max engagement, not for the actual value or functionality.

Enterprise software can be because there isn't an incentive mismatch, good solution is more valuable for the customers, it will sell better and they're willing to pay for it.

But like you say, lot of enterprise software is bad because it's optimized for the payer, not the user, and it's often shoehorned to weird workflows of the particular enterprise.

But your comparison is a bit off; you mention CNC machines and the like to build guitars, but those are tools that are still exactly programmed by humans. LLMs on the other hand are probabilistic - you prompt "write me a set of gcode instructions for a CNC to build a guitar body" and wait / hope.

Sure, LLMs as a tool probably have a place in software development, but the danger lies in high volume, low oversight.

But there's people using it large scale to build large applications, time will tell how they work out in the end. Software engineering is programming over time, and the "over time" for LLM based software engineering hasn't been long enough yet.

I am a craftsman of fine puzzles made from wood and CNC machined metal. I use LLM in lots of ways to help on individual parts of bigger puzzle design projects, like for example to create custom puzzle solver software which can search through large sets of possible notching patterns on wooden sticks in order to find ones that meet some criteria or are optimized in whatever manner I find aesthetically pleasing.

I’ve been writing various single-purpose software tools of these sorts for decades. I would not want to go back to hand-writing them now that I can have agents (cursor, claude code, etc) lay down the algorithmic architecture that I vibe at them, now that I know how to “speak that language” and reliably get the software outcomes that I seek.

I find this similar to how I would not want to spend all day turning the crank handles on a manual milling machine when I can have a CNC mill do it, now that I know how to use various CAM systems well and have the proper equipment.

Given that my overall craft is not limited to just writing code or turning crank handles, I readily embrace any improvements of my workshop “technology stack” so that I can produce higher quality artwork.

I agree. The article's logic is incoherent. It conflates the choice of tools with the decision what product to make and what level of quality to aim for.

If AI can be used to make bad (or good enough) software more cheaply, I have no problem with that. I'm sure we will get a huge amount of bad software. Fine.

But what matters is whether we get more great software as well. I think AI makes that more likely rather than less likely.

Less time will be spent on churning out basic features, integrations and bug fixes. Putting more effort into higher quality or niche features will become economically viable.

I dont think it's the same at all. when weaving was displaced, yes some people were pissed about losing their livelihood, but the quality of the cloth didn't diminish.

when CNC came for machining, no one really bitched, because the computers were just removing the time consuming effort of moving screws by hand.

when computers write code, or screenplays, the quality right now is objectively much worse. that might change, but claims that we're at the point where computers can meaningfully displace that work are pretty weak.

sure that might change.

sighs pulling out this quote again:

"Luddites were not opposed to the use of machines per se (many were skilled operators in the textile industry); they attacked manufacturers who were trying to >>circumvent standard labor practices<< of the time."

Luckily, the brave government's troops, show trials and making '"machine breaking" (i.e. industrial sabotage) a capital crime"' solved the crisis of these awful, entitled workers' demands once and for all and across all time.

I'm sure that any uppity workers in our present age can also be taught the appropriate lessons.

The Luddites indeed lost their jobs to machines, but they could find other jobs, and their children adapted to the changed world.

Dario Amodei, CEO of Anthropic, thinks that this disruption be different from that one. From his article The Adolescence of Technology, currently on HN's front page:

> AI will be capable of a very wide range of human cognitive abilities—perhaps all of them. This is very different from previous technologies like mechanized farming, transportation, or even computers. This will make it harder for people to switch easily from jobs that are displaced to similar jobs that they would be a good fit for. For example, the general intellectual abilities required for entry-level jobs in, say, finance, consulting, and law are fairly similar, even if the specific knowledge is quite different. A technology that disrupted only one of these three would allow employees to switch to the two other close substitutes (or for undergraduates to switch majors). But disrupting all three at once (along with many other similar jobs) may be harder for people to adapt to. Furthermore, it’s not just that most existing jobs will be disrupted. That part has happened before—recall that farming was a huge percentage of employment. But farmers could switch to the relatively similar work of operating factory machines, even though that work hadn’t been common before. By contrast, AI is increasingly matching the general cognitive profile of humans, which means it will also be good at the new jobs that would ordinarily be created in response to the old ones being automated. Another way to say it is that AI isn’t a substitute for specific human jobs but rather a general labor substitute for humans.

It has nothing to do with luddites.

Software quality about speed of delivery and lack of bugs.

If you're fine with software which gets a little bit harder to work on every time you make a change and which might blow up in unexpected ways, AI is totally fine.

Ive yet to meet many AI champions who are explicit about their desire to make that trade off though. Even the ones who downplay software quality arent super happy about the bugs.

> I think one promising shift direction is humans do NOT like to talk to bots, especially not for anything important. It's biological.

Let me tell you why I like shopping from amazon instead of going to a super market...

But also the older I get I keep wanting to visit the store in person. It's not to see the other human, I just want to hold the thing I want to buy and need it immediately instead of waiting. I feel like there isn't enough time anymore.

It helps getting acceptance of talking to bots, when using voice instead of typing book sized prompts into tiny chat windows.

Maybe kids will end up preferring to talk to bots, much like the generations after my own actually preferred digital compression artifacts in their music.

> It's biological.

Nonsense. We never evolved to send text messages and yet here we are with social networks, chat systems and emails used everywhere for everything.