Mastodawn

Stefan Scherfke Feb 7

I think it's interesting how software engineers are (among?) the most eager working class group to replace themselves with LLMs.

It's interesting because LLMs do a worse job than us, we lose ability/skill to do our job the more we use it, lose our jobs, produce worse software, are less satisfied with our work, etc.

Yet so many of my peers seem to be super excited about and advocate for it, while other working class groups at least detest LLMs if not even consider organising themselves to protect their trade/jobs from LLMs.

Are we becoming the cops (read as: class traitors) of this techno-fascist dystopia?

fancysandwiches Feb 7

@vie I'm sure I don't truly know, but I think that it's essentially a form of short sighted eagerness, combined with imposter syndrome. They feel like they're able to do things they could not do before, and they want to be the first/most eager to do it because they feel like that will keep their job secure. But really, if your whole job can be automated with an LLM what incentive is there to keep you? What value are you providing if you don't understand what the code being produced does?

☭. evie Feb 7

@fancysandwiches I agree, that's absolutely a part of it. But I think it's also feeling superior and smart while being absolutely fooled by these billionaires' marketing to the extent that they call it "AI" instead of "LLM" and believe in magical claims from these companies instead of math.

It's probably this contradiction is key to short-circuiting thought and just stopping further thought?

fancysandwiches Feb 7

@vie oh for sure there's a level of feeling superior to others mixed into this. I reviewed some code this past week that was really bad, super backwards, and nothing like anything I'd seen before, and I said as much in my review, but more diplomatically. I was then lectured about how it's actually really clean, and a "well established" pattern. The problem is that the pattern they said they were using wasn't what was in the code, and my suggestion for what they should do, was actually the pattern they were referencing. They had no idea what they were doing, yet they felt it necessary to tell me that it was well established, and "clean", and that I was wrong.

You're hanging out with a different bunch of software engineers than I am.

I have to admit, some sw engineers are horrified by the applications that business people want to use generative AI for, and some are keen.

But even the podcasters that talk about AI are saying it is not ready for prime time. Then they narrate or play a rah-rah ad. It's quite disorienting. But they need the money I guess.

Chris [list of emoji]Feb 7

Most of my peers, when they use AI, treat it as a power tool. None of them think it will do their job; it's more about whether or not it will make them more productive in the same way that an IDE with refactoring tools would.

(IM reluctant E, maybe a little bit in some very narrow cases *if* you're careful and understand the problem well.)

Antoine Leblanc

@vie i wonder why that's the case, but yeah i've observed that too. my gut feeling is that code is extremely predictable, so LLMs are better at that than they are at other things, making them more enticing? and then the chatbot psychosis sets in.

☭. evie Feb 7

@nicuveo Yeah, I think the fact that they tend to produce verbose and large quantities of code hinders our ability to judge it as harsh as it deserves. And in some ways, it's impressive. Or it would be, if it wasn't used to throw code to prod :)

I've had a theory for a few years that most software engineers don't actually like software engineering. Had there not been money in it they would have followed a different career path instead of getting a CS degree from a 4-year college (or boot camps).

LLMs align with this theory. The people who are excited that something else is doing their job for them are the same people who picked the job for the salary, not the joy.

The Other Brook Feb 7

@jason @vie It used to be people who were into computers got CS degrees and people who were into money got finance degrees. About 20 years ago CS programs began turning out people who were into money, not computers and the outcomes have been terrible.

@jason @vie @theotherbrook When I started my CS degree ~30 years ago (Which I did because I enjoy programming and working with computers), the first year was huge due to all the people seeing programming as good money. Most of those people didn't make it to into second year, and even fewer into third year.

My understanding is the intake started to drop shortly after I finished.

@puck @jason @vie @theotherbrook timing is not quite right on "20 years ago": that was 2006 and enrollments were still pretty tiny post dotcom bust. (source: I was on the faculty job market in 2006.) Maybe 10-15 years ago is closer.

The Other Brook Feb 9

@va2lam @puck @jason @vie 20 wasn’t intended to be a precise number, but based on my experience going into a non-CS master’s program in 2008, I think that year a lot of people pivoted to tech and tech-adjacent fields because finance was in a shambles.

@theotherbrook @puck @jason @vie 2008 might be about right.

Chris Hanson Feb 9

@va2lam @theotherbrook @puck @jason @vie It may have accelerated or resumed then, but it was also definitely a thing during the mid-late 1990s dot-com boom.

(Obligatory: “I was there, Gandalf, three thousand years ago…”)

@eschaton @theotherbrook @puck @jason @vie ... always cyclical ...

Chris Hanson Feb 9

@puck @jason @vie @theotherbrook Unfortunately the universities’ “solution” to this was to make the first year of the standard computer science course less difficult, rather than to try to figure out how to better direct students to the appropriate degree.

I say this as someone who dropped out of a top-tier undergrad CS program for academic reasons in the mid-1990s: I strongly suspect I’d have done just fine today, and that’s not actually good.

Wolf480pl Feb 9

@eschaton
huh, I started my CS degree in 2015 and the first year was difficult, and the dean was like

"If you fail the first year, that's by design. It saves time for both you and us.
If you fail the second or third year, that means we fucked up, and made you waste years of your life on sth that won't work for you."

@puck @jason @vie @theotherbrook

rollspelosofen Feb 9

@puck @jason @vie @theotherbrook Sounds like my program. When I started my degree we were about 70 students and after the third year later we were about 20.

BenderIsGreat34 Feb 7

@jason @airshipper I came to the same conclusion. I know so many who went into software as a path to management as well, and they were terrible programmers and only needed to hang on for a couple of years before switching internal roles. If AI coding had been a thing 20 years ago they would have been all over it from day one.

Recovered Expert Feb 15

@andymoose it’s more than a theory for me. Ive seen many since the late 1990s and they’ve often been quiet open about the reason being paid very well *and* (seemingly) not having to word „hard“ for that. When the dot com bubble blew it slowed down a bit instead of dropping to 0 and the the next hype cycles came.
Many w/ technical jobs were eager to go into non-technical positions, incl „consulting“ or mgmt.
Not all Managers are like that, but the % increased.
@jason @airshipper

Chris Hanson Feb 9

@jason @vie Unfortunately I know some excellent engineers who have gone all-in on LLM slop. It really made me re-examine my position, but I wound up at the same conclusion: They’re just—and very unfortunately for them—wrong. It sucks to know because I really respected them, too.

fluffydotorg Feb 9

@jason @vie heck yeah. When people do LLM demos and talk how great it is they don't have to problem solve or "type stuff" any more I'm like "why did you sign up for this career?"

Bennolius 😷⚡Feb 9

@vie @jason could also be that they got into it because of joy but it's been sucked right out of them, so now they just want to get paid for not doing any work. (I haven't found much joy in my work for the last ~10 years, but I still want to do good work and not offload my thinking to a magical plagiarism machine)

Dominykas Blyžė Feb 9

@jason @vie not a very good theory

I'm super excited about being able to build things that I couldn't before. Also with career progression, you start to appreciate how much of the building happens under the surface. The quest for Building Properly, for me, has resulted in quite some builder's block. That has now been simplified.

Wolf480pl Feb 9

@jason
What if you went there for the joy, but ended up in an understaffed company, with low bar for quality, and discovered that sucks the joy out of it?

☭. evie Feb 9

@wolf480pl @jason If I end up truly believing that programmers are the same as cops, I'd probably quit the field.

It's the same principle which I use when talking about ICE for example. I don't care if they need the job, a line in the sand needs to exist somewhere no matter how much we agree with the concept of no ethical <anything> under capitalism.

@vie @jason The way meetings are going at work recently I sometimes get the impression I'm the only one who cares about developing stuff, there are quite a few people who seem to be desperate for Copilot or Codex to do absolutely everything :-(

RichardW Feb 12

@vie And now the powers-that-be at work have declared a goal of having at least 90% of all development done by Codex, so, yes, awesome

Andy Wootton Feb 9

@vie @jason There is another way to look at this. Most writers hate writing. Many are highly motivated but constantly disappointed in themselves. That's perfectionism. We want that. Being rushed to release something faster and worse than they want to is painful too.

SOwOphie ⠬⠆Feb 10

@vie @jason An LLM power-gluesniffer once told me something along the lines of "LLMs are perfect for code that you don't care about."

I fully agree.

it's winter Feb 13

@jason @vie well at least it cured my imposter syndrome
I do not want to be replaced
I want to make software because I like making software
I am, in fact, a real software engineer

BradleyKSherman Feb 7

@vie Computer Maoism. It's endemic in Silicon Valley.

Chris Hanson Feb 9

@bks @vie Now, I’m not a “theory” guy, but I’m pretty sure Computer Maoism wouldn’t endorse rent-seeking…

☭. evie Feb 9

@eschaton @bks Digital Maoism is using "Maoism" in the capitalist boot-licking sense, by looking down at any collaborative community effort, exaggerating its flaws and downgrading its benefits.

In a way, if we use this definition, it is true. It is stealing the collective community works (open source software) and creating a shitty (computer) product out of it.

But I object to using "Maoism" derogatorily when rampart capitalism and techno-fascism will do just fine.

BradleyKSherman Feb 9

@vie @eschaton It's the computer as dictator, and the dictator's every pronouncement as doctrine. Not an economic theory. I'm guessing neither of you are old enough to have actually argued with a Maoist.

Andre Videla Feb 7

@vie I feel like LLM only exacerbate a situation that already existed: There are artisans, and there are workers.

The artisans build things out of their craft, and the outcome benefits from decades of accumulated experience.

The workers perform a task at hand, moving boxes from A to B, mainly worried about the outcome of not having box A not reach B in time. Speed and compliance is key in a competitive environment like that one.

It's always been hard for artisans to find work in an industry hungry for workers. And it's been tough for workers to outcompete each other in speed and price.

Coding agents provide workers with the unprecedented ability to clone themselves, sometimes many times, and get many boxes moved from point A to point B. Because what they sell is task completion and not craftsmanship it doen't matter in what way the work is done. And the industry never cared in the first place.

There might be a world where agent orchestration will become a form of craft, and the ones diving deep right now will have a head start in this industry. Or maybe not and it's all just hype, smoke and mirror. there is no way to tell.

I'm happy with my craft, even though it's never been recognised as such and never will. I wish good luck to those surrendering their skills to the whims of large undemocratic corporations.

Badly-read literary snob Feb 7

@Andrev @vie I like this analogy a lot. Someone else told me about being a poet vs a journalist and although journalists have screw it up the last decades I thought it was unfair to journalism.

Anyways, it also reminded me of my experience in university. I was a bad architecture student, but I was fascinated by theoretical subjects, my classmates that were good at it often didn’t care that much about the theoretical aspect and wanted to shorten the career by removing them from the curriculum. We had a discussion once when they said «this is not useful for work» and I was like «I’m such a privileged bitch that I never considered working, I can solve that later». I had a romantic idea of what university was, as well as of programming.

Is nice to know I’m not the only one, but I don’t just wanna carry the feeling. I feel like we can still do something

Chris Hanson Feb 9

@RosaCtrl @Andrev @vie Remind yourself that if not for people like you, they would not be at this place. And that’s not a negative: Someone has to push things forward to prevent stagnation. But it really fucking sucks when it’s not acknowledged, recognized, and rewarded.

Cap Ybarra Feb 7

@vie huge ratio of opportunists in the field who never liked the job to begin with and just were looking to cash out. they would have been in finance in the 80s but tech is where the money is today

Cap Ybarra Feb 7

@vie just been more and more tourists every year in the industry. gish gallop for the recruiting world which is why that has also gotten progressively worse. hiring is a nightmare now

@vie create something new. "It" can't.

RootWyrm 🇺🇦

@vie software engineers - especially in startup culture - are arguably the most dangerously delusional idiots in the history of labor. (Not helped by people insisting they are overpaid.) Which has only gotten worse. They are absolutely convinced they are just temporarily embarrassed millionaires whose only ally is management and the VCs.
And now they are just one prompt away from their killer app that will make them a billionaire. For sure.

RootWyrm 🇺🇦

@vie @davidgerard and some of this is driven by the myth of the dot-bomb millionaires. You know. Those employees that management gave a bunch of stock and then they retired at 30. And you know what the actual result of that was?

Even before the dot-bombs imploded, management made DAMN sure that could never ever happen again. But you will still continually hear idiots at pre-IPOs insist on offering day, they're going to get to retire to Hawaii. And a union would take that away!

thezerobit Feb 7

@vie
Every time something dumb happens in software engineering, I feel like my value goes up. Other people losing braincells from LLMs? More braincells for me. I think that's how it works. Among my coworkers, very few seem to be advocates so much as cautious adopters. Not a lot of class consciousness or solidarity in this industry.

Steppe Lord Feb 7

@vie tech workers are probably the sector of the economy with the most temporarily embarrassed millionaires, of course most of them would hold the tech in high regard, if not for coping with the usual phrase "AI won't replace you, a programmer using AI will".

As a "builder" "indie hacker" with "entrepreneurial" mindset it must be hard to deal with the cognitive dissonance of both believing this tool you use is so good it will replace large swaths of the workforce but also coming to terms that the most work you do is prompting and that you too, may be as "expendable" as the other people.

Also, this is biased as I avoid reading xitter, but I've seen most boosters are wealthy programmers, i.e. they worked in the sector when software engineers were the darlings of the working class, and made their riches when they were seen as indispensable. Of course 10 years later they're not worried about juniors or the future of the profession, they have their life solved.

saruman mentality

David Mankins Feb 7

The hand-tool artisans look down on the users of power tools, with some justice. But it could be that the two sets of people have different goals.

For the artisans, it is the process itself. The constructed artifact exists primarily to embody the care and skill that went into its construction.

For the power-tools user, perhaps their goal is a shelf to store the books they were able to read because of the time they did not spend in the workshop.

Alex Wilson Feb 7

@vie @lain_7 The artisan knows a good tool when they see one and if it were useful and truely labor saving they'd be using it. Poor metaphore that presupposes utility and ignores the critism of a group of experts.

David Mankins Feb 7

You’re missing the point: all these artisanal-based arguments, which I’ll summarize as: “you’re not really interested in being a software engineer”, seem to be based on the assumption that being a software engineer is the only goal one might have.

There are other goals, for which programming is the means, not the end. A bicycle for the mind, remember?

My own experience with these tools are that they lower the activation barrier on undertaking a project. I achieve in an afternoon what would have taken several days. The effect is that of having a really good library of functions that are geared to the task at hand (“import antigravity”: https://xkcd.com/353/).

But, you know, 90% of everything is crap. Making it easier to make more things brings with it the risk of making more crap as well.

When desktop publishing was introduced, graphic designers found a lot that horrified them.

When the web was introduced, we all got really familiar with the horrors of the <blink> tag.

I’m not saying one should ignore the criticisms of experts, instead take them to heart in using the tools mindfully.

Python

xkcd

Alex Wilson Feb 7

@vie @lain_7 I love tools that make programming accessible and easier. I love piles of people's terrible scripts that work just for them. That's the beautiful part of computing.

I don't love an industry hell bent on renting skills that should be free. LLM output being 'bad but functional' is a feature in keeping people from learning and weaning off the system. The industry is for a profit so models are tuned not to help but to keep you coming back for more. That's what I talk about with it being bad.

They're buying up all the ram, gpus, and storage. Personal computing is being assaulted on all fronts with the aim to sell it back to you every month. After this industry implodes maybe there will be a use case for some llm tools but for the moment they're the weapon of a dangerous foe and helping them push it harms us all.

David Mankins Feb 8

I think you need to separate the tool from the hardware store.

For example, there are an increasing number of models that one can self-host. My employer will only permit the use of self-hosted models for confidentiality reasons (confidentiality of their own, and their customers’ data).

For personal use — my computers tend to be scavenged from remainder sales and dumpster-diving (found a marvelous 17-inch laptop in an e-waste bin the other day, only ten years old!) — I’ve chosen to rent from Anthropic (a single new PC purchase would cover five to ten years of subscription fees).

Yes, you can use these tools wrong — a thumb and a board look much the same to a consumer table saw; measure thrice cut twice is an easy habit to slip into — but that’s up to the carpenter, not the tools.

🏳️‍🌈 Bri 🌧️🚴💦✨Feb 7

@vie Essentially yes. I might argue that the overall industry in which software developers work functions to undermine labour power, and has for long time ☹️

I'm a former software developer and I left in part because of my discomfort with the lack of thought or support offered to those whose jobs get automated. I think the industry selects for a specific individualistic mindset, whether intentionally or not, and I don't find the lack of solidarity or caution surprising 😮‍💨

@vie well the "senior" ones usually think only the juniors will get replaced... Thus as a bonus they won't have to deal with any juniors... As if any management wants to keep the most highly paid around to babysit their BS generators.

Bogdan Buduroiu Feb 7

@vie most modern software engineering is thin layers of logic plumbing together different SaaS providers. Your boss will never let you self-host Postgres because he thinks it's cheaper to get an SLA from GCP. It's boring, repetitive, soulless. I'm not surprised people would be eager to turn it into an assembly line of PR reviews instead of writing code.

Also, engineers are very competitive, and most think of themselves as special, that's why unionisation has been historically low in SWE. They're not thinking "AI will take my job", they're thinking "how can I use this LLM to outperform my colleague, get him fired, and get his bonus".

Dustin Metzgar Feb 7

@vie I'm no fan. But years ago we had the whole copy-paste from Stack Overflow as the bad habit of SEs. I think of LLMs as slightly better at copy-paste. I use it in small quantities and fix what it gives me. It saves me time. And it's helpful for researching brownfield codebases. It lies, convincingly, but it gives me an attack angle.

I do get unreasonably upset at people using AI to generate documentation and then just checking it in without reviewing it. That's almost worse than no docs.

Julian Gonggrijp Feb 7

@vie I think part of the logic is that a lot of developers have low self-confidence. If you believe that you're not as good as your peers, it seems attractive to defer part of the work to a machine that presents all its answers with certainty. It lessens your perceived burden and you can lean on its apparent confidence. It's actually bluffing away, an unreliable source compared to intentionally written documentation and a much worse programmer than the nonconfident human, but when you lack confidence, this is harder to recognize.

It's a sad vicious cycle where the nonconfident becomes dependent and even less confident as a result. However, I think we need to recognize these people as victims of a trap rather than blaming them for it. It's mainly the marketing rhetoric of the genAI vendors that is at fault. The "mystery AI hype theater", as @emilymbender aptly calls it.

Tip for all humans: "On Confidence" from The School of Life is a sweet, smart, funny, heart-warming little book that you can read in an hour. It might also boost your confidence a bit. Buy multiple copies, give it away to friends and colleagues, bring it to work, on your travels and to the toilet, read it multiple times. One of the best books I've read.

https://www.theschooloflife.com/shop/tsol-press-on-confidence/

#genAI #programming #confidence #SchoolOfLife

On Confidence Book | The School of Life

On Confidence - This book walks us around the key issues that stop us from making more of our potential. The School of Life.

The School of Life

Jean-Sébastien Guay Feb 7

I think you probably see another crowd than I do, or I'm in a bubble that resembles my own views, or something. Because what I see (and my own position) is that software engineers of experience recognize them as tech debt accelerators and a massive security risk, and want as little to do with them as possible.

For my part, I am already saying that if things really do go the way of more LLMs everywhere, I will be considering a very early retirement. I do not want to pick up that mess.

@vie It's worth noticing that the ones most enthusiastic about it are also the ones that care the least about their craft and were only ever in it because it was a well-paying job (in some places, they wouldn't have chosen it otherwise).

That's not a coincidence.

Cassandrich Feb 7

@lispi314 @vie This this this 👆 💯

They're also the ones who engage in all sorts of fascist-aligned work other than AI, because that's where the big money for low competence is.

The pay is for being a fungible cog in the machine, not for any high level of skill or understanding. They compensate for that in numbers and willingness for lots of their endeavors to be dead ends. They have plenty of money to invest in making countless trial variants of the torment nexus.