Clojurians Log v2

Of all the discussions I saw about this, not one comment on how privileged and entitled this behaviour is. Oh yeah I'm going to be disruptive and cause problems because I'm Right ™️. I'm a professor don't you know.

There will be some friction initially with delivery robots until we figure it out but I can find no reasonable objections to them

I can't overemphasize this - we can't have a society with people who believe being right justifies being anti social - that's a PVP server

It's not a large leap in my opinion, when you create a permission structure, etc. Anti social behavior does not distinguish the target, that's the point. When you create a moral license for yourself to behave a certain way, you also habituate this mode of behavior. Combined with "scope creep", it will move up from delivery bots to "people I don't like". Not that it hasn't happened already. My silliness is in trying to close the barn doors after the horses have all fled.

Setting aside the HL reference, it would be more accurate to compare it with intentional littering, the problem with this logic is it cuts both ways. I don't want to live in a pvp server where the streets are dominated by two opposing groups who are sure THEY are in the right and engaged in RIGHTFUL non compliance, which is the degenerate end state of this approach. This only works when one side is reasonable - in a game of Chicken the least reasonable party wins. I'd rather things don't get reduced to this state.

The problem with this approach is that once a group has convinced itself / worked itself up to / worked up the courage to believe they're living in a situation like the one you've illustrated (ad absurdum), it creates the exact moral license to use indiscriminate violence regardless of whether it's TRUE. As external observers, it is only justified if we agree with them. Curious how that works.

I don't want to live in a pvp server where the streets are dominated by two opposing groups who are sure THEY are in the right and engaged in RIGHTFUL non compliance

I HATE THE CONCEPT OF THE POLITICAL

I'm not a historian (far from it), but it seems to me that the argument "your form of protest is wrong/counterproductive" is historically commonly used by the powerful when they understand that they have run out of real (ethical/logical/humanitarian) arguments.

I'll note this methodology of moral licensing was also used by every historical violent power grab that ended in mass graves, executions, and other fun terrors.

It's not "your form of protest is wrong", it's "you're participating in creating a moral permission structure that ends up in violence, please stop"

Personally I'm not going to engage in any meta-protest-form-debate. IME such debates tend to remove focus from more important issues.

Sure

That would deprive jobs from the slave labor we imported for reasons

Will the freed up resources from such automation go to "imported slave labor"? Oh, right.

Phrases like "create a permission structure" and "moral licensing" can ignore the fact that there are already existing permission structures and moral licenses. We're already swimming in them, all too often reflexively.

Like I said, these horses have fled the stable long ago

Thanks for the link. I've found Kyle's other writing in this area worthwhile as well. https://aphyr.com/posts/411-the-future-of-everything-is-lies-i-guess

Calling attention to this statement: > Where correctness matters—and surely it ought to matter for databases!—our industry must acknowledge and mitigate these tradeoffs. This correctness tradeoff isn't new in software (or anything else, for that matter) with in introduction of LLMs, and is something I struggle with. A slight twist on this is "when is something good enough?" I know I lean much more towards correctness matters. I want to tools that I can trust, and that I enjoy using, and part of that enjoyment comes from the quality of the tool. I know others lean towards "if it gets the job done, it's good enough", and that gets supercharged when "can we maximize short-term profit on it, it's good enough" comes into the picture.

> More broadly, the use of LLMs erodes trust in documentation, in software, and in the organizations which produce them. Emphasis mine. I would like us to push even further on this, to the individuals who are part of such organizations. There's a tendency for people to shed their own personal sense of culpability when they're part of an organization.

There are arguments to be made (and will be) that Orkes wasn't using LLMs correctly, that they didn't have adequate specs, that advances in LLMs will make these types of issues less so in the future, and the inevitable "these problems aren't new". When I raise the question "is it worth it?" some will answer "yes". And in a sense these all come down to value judgements. What do you value? How do. you prioritize your values? These are value judgements.

The first comments on a thread can often determine the direction it goes. I'm not sure if this particular direction is going to be the most useful, but these are the types of things I've been thinking about. If others have other directions they'd like to take this, I won't hold it against you.

I think it's often an uphill battle to act on your values.

I think we want a society where its not so much of an uphill battle to live our values.

(And thanks for the reminder from Richard Gabriel!)

Phrased even a bit more forcefully, do we want a world where the psychopaths have inordinate power? Where the psychopaths "win"?

Calling what we have a free market cedes far too much to rent-seeking technofeudalists.

Agreed. Though those self-described "pro business" arguing for less shaping of the market tend to insist on that language, and I (probably mistakenly) try to start from some shared definition of terms. It's an attempt to avoid getting dragged into a "no true Scotsman" type of debate. (Due to the vast gaps in my economic history understanding, I recently learned that Adam Smith himself had unkind things to say about rentiers.)

I haven't learned/developed an alternative framing that isn't confused with command-economy socialism. Reading suggestions (sans astrological influence, please) welcome!

You could cut it off at rent seeking and still be correct

You could read some of the libertarian literature, which ranges from quaint (I, Pencil) to rigorous theory (Rothbard). It's nice, logical, ordered, optimistic, and if it feels slightly detached or not applicable for reality, too bad for reality. Alternatively, you could dive into more modern value neutral analysis, like that of Complex Adaptive Systems, which is a cool field and the Santa Fe Institute does good work. Finally, you could accelerate into the astrology wall and break through to the other side and read some Nick Land, who, while out there, presents an intepretation of techno capital I've been unable to refute. Don't take anything he says as normative, but it's certainly an alternative (and strong!) framing.

Just thinking out loud here... Modern cars are quite safe, but manufacturers could add even more/better safety features and we could also cut all speed limits by half. However, more/better safety features would increase costs and therefore the final price more than consumers would accept and the same with the speed limits - would likely upset the voters. Similarly, the quality of you average SaaS app, e.g. a CRM or whatever is a tradeoff that consumers "choose." If you're bringing a product to market, you'll lose out to competitors if the amount of effort you're putting into your app significantly exceeds what consumers demand because you're competitors will deliver the desired tradeoff faster/at lower cost. We're assuming for the sake of discussion the competitors are equally good at software engineering, so the differences boil down to effort invested. Of course, different apps have different demands for quality, and there are different consumer segments. But within a category and segment... Still one wants to do the best possible work as a matter of pride, but there are market contraints.

So for me, I use LLMs to generate code, but I then review and edit every line of code. I also use LLMs in other ways. And some parts I hand code. The key is to stand behind your work, however you get there.

Agreed that there are trade-offs. For the sake of the above discussion, I think we're also ignoring cost of entering a market, and what is available to end-consumers for purchase. I know there are things I would like to purchase (and willing to pay for) but aren't actually available, and I can't afford the start-up costs to actually make it myself. Similarly there are things I used to be able to buy and am willing to pay for, but are no longer available because they were no longer sufficiently profitable and discontinued and/or bought out by competitors or private equity and shut down, or what have you.

> So for me, I use LLMs to generate code, but I then review and edit every line of code. I also use LLMs in other ways. And some parts I hand code. The key is to stand behind your work, however you get there. I think one of the problems that gets lost is the sense that the only thing that matters is the code that gets checked into version control. No other other direct/indirect benefits or costs matter. The car example is illustrative. Should we be building communities around roads and cars? How would we design cars for walkable cities? Should we be designing cars and roads that maximize the total number of cars and miles driven?

I think one of the problems that gets lost is the sense that the only thing that matters is the code that gets checked into version control.

The wording could be improved (I should have finished my coffee before posting). Most LLM discussion focuses just on the code that gets checked into version control. Other direct/indirect costs and benefits are ignored. The problem that often gets overlooked is "what are the likely long-term consequences of building software with llms?" which provides a broader context than only considering the code that gets checked into version control.

I think that matches my desired reading. What's lost is the larger context, when focussing just on the commits. The idea that "the only thing that matters is the commits" is not what gets lost.

BTW, these jepsen analysis and associated Clojure workloads are pretty awesome and clean/easy to read. And besides their serious work we all know well, they also make good jokes (screenshot from their previous article, also for the Orkes company)

"a fractal of correctness errors" 👈 I'm stealing that one

Let's say you’re close to an expert climber. A fellow clojurian who's been climbing a year uses AI to write a book to their liking. Would you be interested in at least looking at it? As a close to an expert climber I sure would.

If 1000 monkeys with typewriters write a book, are you curious what the book says, despite your limited time?

I say that lightheartedly, and don't advocate that we invest money into monkey stenography lessons.

Also I do admit your scenario implies "nontrivial decisions" which is a fair critique against the monkey book argument.

But we would never even know the monkeys succeeded if nobody took the time to read it and say, "Yeah that was surprisingly profound and error free!"

I'm actually also wondering "At what point of excellence would you consider begin reading a book written by the monkey legion?" If the book is a masterpiece, and everyone is saying how amazing the book is; is it automatically invalidated via the process through which it was manifested? Not trying to be petulant, but I wonder if there would ever be a point of perfection where such a wonder is created, but the means of creation nullify whatever was produced? Is there an ontological difference between hello world made by 12 year old us and the first llm generated hello world?

the astrology book quality only sucks because you aren't using the right harness, obviously

I play a lot of TTRPGs. Part of making them work is consulting an oracle (the dice). Some of the games are about creating little stories. Is there a difference between consulting the oracle via a table and some polyhedral plastic bits and via a bot chat? Are some stories made “better”?

Re: the ontological difference between a twelve year old’s hello world and an extruded one: that answer depends on one’s conception of what it means to be!

2, I wildly presume the astrology-assisted book has more content, which content by stipulation is expert-reviewed and -validated, We also have as a given that the human expert (well, another speculation) wrote the book, meaning I won't have to try and fail to read impenetrable AI techspeak. Now if you had said ouija board...

Someone here had recently directed claude to create a book about indoor climbing. It made me think of the book “Mountaineering: The Freedom of the Hills.” That book has descriptions of facts and procedures, but also knots, integrated techniques and under which conditions to choose them, situational awareness, etc… I would only choose #1. I would hope that the expert writer would have fact-checkers. I would be very concerned that while the stars may mostly align in the astrology cases, if they didn’t and the expert didn’t catch it (perhaps through star-blindness?). Were I to rely on the latter two books, say, to read a snowpack’s profile, and it was conveyed wrong or just had the wrong facts, it could be fatal. I’m conservative in this way though.

I'm not sure why s/LLM/astrology/ in OP. To me it signals the author already decided what the correct answer is, undermining any attempt for a constructive discussion. > skeptic: Someone who habitually doubts beliefs and claims presented as accepted by others, requiring strong evidence before accepting any belief or claim.

if the "author" didn't bother to write the book, I'm not bothering to read it... so 1. for me if I want an AI to tell me about a subject, I'll ask it directly

That's interesting. So the input of the person guiding the LLM has no value to you? To put it in coding terms, to you there is no difference in the code the LLM outputs, regardless of whether it was me (35 years of coding experience) or my daughter (3 weeks of coding experience)?

we are talking about a book here, not code

the question that was posed, was given these 3 choices and limited time... I will always take the book written by a human expert

So let's bing it back to the concrete example then. @simongray, having climbed for about a year and having read several books on climbing, not finding what he needed, guided an AI to write the book he wanted. So basically what you’re saying is that the research he did, and how he guided the LLMs have no value and you could have done the exact same thing?

I could imagine a scenario where an AI generated book would be ok, but not this one

where did I say it has "no value"? but I would use my time reading a book written traditionally by a human expert

You’re right, but you did write “if the “author” didn’t bother to write the book, I’m not bothering to read it” which I equated to you saying that the “author”’s experience and guiding of the LLM had no value.

sure, I can admit that much... given there's human written books on any subject I care about, I would always go for those first.

I guess it is a coming reality that there will be more AI generated (or assisted) books coming out and the markets will decide... but the audience doesn't owe authors their time or money, if they don't like AI generated stuff, that's their prerogative

I think there's an important thread that I brought up in the #C068E9L5M2Q thread about this, which is that the book in question here is a free and open source text. I think if an author is selling something then there is a requirement for thoroughness that I think is reasonable to request of an author. It would change the context quite considerably. My take is that it could still be defensible to use LLMs even in the example of someone selling something, but I think they should have ensured the code does what they want everywhere and that they can justify it to the people paying them; and if those people inquiring are not satisfied then they would deserve their money back and if not honored then said author could be said to be promoting a scam.

it is perfectly fine imo to use AI tools to make a book you want to make... but it is also perfectly fine for others not to want to read it.

if you ask people to consider something you made, you are in fact "selling it"... even if money doesn't exchange hands, but because people can choose to spend their valuable time in a myriad of other ways as well

so I think not disclosing that it's AI generated, you are not valuing the recipients time and that's rude

@xifi exaggeration for the sake of illustration. Also, skeptics can have strong beliefs, as long as they are open to the possibility of being wrong. I too would like to see #C0B7Q4XRM1U be less of an echo chamber, and I think it can be. I think we might be surprised by how good those discussions can be, but I think that would require the AI positive people to start the conversation.

@tatut I didn't think my response precluded a lack of disclosure. I think that authors should disclose how much and the way in which they utilize ai if they do. True that you are philosophically selling any idea, but isn't the whole point of a marketplace that you can take the business where you prefer? I didn't insist anyone should read LLM generated content but rather that people shouldn't be stigmatized against it because it doesn't appeal to everyone. Trashy romance novels are garbage and destroy many trees a year, I think we should go after them instead, and the current executive administration of the US but that's another story. Edit: thank you @soltanzadehramin for catching my tagging error

I might also add that I actually consider myself an ai skeptic even despite using it. I detest LLM writing (in fact as soon as someone says they use ai for something, I am immediately skeptical and have many questions) and I think LLMs are obviously tasteless and produce garbage output 99% of the time. I just think they could also be used in novel ways and to jump on someone for trying something new is a bit counterproductive.

@wistfulmelodylispfulp I think you meant to tag @tatut. Right?

My bad you are correct, just on mobile and intermittently responding here while working. I promise I didn't have ai do it, maybe it would have done it right.

@soltanzadehramin as it stands I'd just rename it to #ai-haters. > if you ask people to consider something you made, you are in fact "selling it" too strong wording for my taste. People release open source software, with no expectation of any profit, just to share. If you mean "promoting", OK. > exaggeration for the sake of illustration Why are we exaggerating one side only? Overall, the tone here is far from skeptic. I am skeptic towards LLMs in some ways, but this channel is, as you called it, an echo chamber.

I use AI tools as well, but code (even Clojure) has lots of things structure that is there just for the machine... and repeats in the same way very often. I think a book is more purely about human to human communication, so I want the author to write it with the audience in mind... not to just generate it. does that make sense?

also regarding disclosure, if the AI outputs are good enough that you deem them worthy to be included in your product, then proudly display it instead of hiding it

@xifi I think that's usually how it goes with metaphors in debates. We choose metaphors to illustrate our POV. To someone who strongly disagrees that metaphor is likely to sound overly simplified and close-minded even though it was likely largely intended to simplify the discussion by reducing it to one dimension. Think of the classic "You wouldn't steal a car." advertisement against pirating.

I agree about #ai-haters, but that seems to me to be a result of non-haters staying away or mostly lurking rather than it being something in our control.

FWIW, I consider myself both a hater and a skeptic. I doubt I'd be alone in that.

hater, skeptic and user here

> I'm not sure why s/LLM/astrology/ in OP. To me it signals the author already decided what the correct answer is, undermining any attempt for a constructive discussion. @xifi I asked a question. I invite you to play along by trying to answer, or asking a counter-question, suggesting problems with the analogy, or suggesting alternative analogies. Possibly in a new thread as this one has gone in many directions.

I read it as a https://en.wikipedia.org/wiki/Loaded_question. I don't wish to accept astrology = LLM, which your question forces me to. I don't want to be invited to play a rigged game.

As I (just) said, I welcome criticisms of the analogy. By all means, don't accept it. I explicitly offer you the https://en.wikipedia.org/wiki/Principle_of_charity, and I ask you to reciprocate.

imo astrology is like the inverse of using an LLM - the position of the planets give the experts prompts to consider and then allow them to consider perspectives they may not have. I would gladly read an astrology inspired book but not an llm generated one

LLMs in my experience shut down my thinking process

I'm not a climber but if I were to read or refer to eg a recipe book I'd want the content to be human written and tested. I wouldn't object to automation for layout etc

We got #astrology-assisted-coding before GTA 6!

I noticed that as my expertise started to grow over time many people flooded the internet with articles like your point #3 before AI was a thing. I personally have a disclaimer on my blog portion of my website that states that LLM usage is not used for anything unless explicitly stated because that just feels courteous to me. Some of my more recent posts could have benefited from some LLM review, but I decided I'll just leave the posts with their flaws and correct them manually later on if they embarrass me too much.

If I ever see the phrase "load-bearing" in anything I'd immediately stop reading.

I sensed prejudice in the question, that was my criticism. I am OK to be proven wrong, my reading is, by definition, also subjective.

> imo astrology is like the inverse of using an LLM - the position of the planets give the experts prompts to consider @danieleneal interesting. Could not an LLM also be consideration-triggering? I certainly have tried to use them for that purpose. I suspect that one big difference is the friction. Astrology doesn't take over as much of the job, leaving me wrestling with the material. With an LLM, cognitive surrender is too easy. "Consider" is a great word for someone's future conference slide. Cognitive surrender is literally in-considerate.

@stuarthalloway Ahaha yes. Consider indeed is a brill one :) Glad you looked that up - and reminded me also

Yes exactly. LLM leads to cognitive surrender. Asking the stars usually leads to more questions, possible tensions to resolve I liked @fogus 's piece on this and what he would want from an AI assistant - for it to be more like a socratic questioner https://blog.fogus.me/meta/LLMe.html I guess you could make them behave like that now? Perhaps? But it is far from the default...

Incidentally I recently finished writing an album using astrology to make non trivial decisions https://untitled.stream/library/project/a1lCVrP1zPK56RXAamGu8 ;) https://incorruptiblebodies.com/products/the-sun-the-moon-and-the-wandering-stars/

Well I think it's easy to be dismissive and say that

(= cognitive-surrender (LLM (or usage  implementation)))

@wistfulmelodylispfulp It is my experience in personal projects that the more I use an LLM to generate code the more averse I become to working directly with that code, and the barrier to entry seems to increase. Is your experience different from that?

Of course I do experience similar things with co-worker code, but I seem to be able to overcome this by asking for a walkthrough from the author

I think cognitive surrender is a (probable|likely|potential|unlikely) side effect of LLM use. We will see which tier of probability it is.

In the same way that building expertise was a (probable|likely|potential|unlikely) side effect of designing a system by hand.

@wistfulmelodylispfulp No slippery slopes here. I used LLM chatbots intensively for six months and never surrendered! But everyone thinks that about themselves, so I also looked at studies: > Our main finding is that using AI to complete tasks that require a new skill (i.e., knowledge of a new Python library) reduces skill formation. -- https://arxiv.org/abs/2601.20245 > Using large-scale evidence from a natural school setting, we Önd that generative AI use substantially reduces learning among secondary-school students. The central pattern is stark: AI raises homework scores and reduces homework time, but lowers perfor- mance on closed-book exams. -- https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6868618

Applying charity, I don't see how this could be considered as such as it was a direct reaction to Simon's post. Grouping LLMs and astrology should be considered category error. I don't have to tell you LLMs work, they're pretty good at predicting the next token, while astrology is persistently bad at making any prediction. Thus, when composing a string of tokens, I won't dismiss a book generated with LLM assistance on its face without considering both author and process. The findings you quote are undeniable and I myself have predicted this about a year ago, search in Clojurians slack "auto amputation"

> I used LLM chatbots intensively for six months @stuarthalloway - I'm curious if you used agents (vs chatbots) at all?

@ben.sless in my experience astrology is less for making predictions as it is for subjective inquiry into the quality of time, a bit like tarot. See - kairos vs chronos https://en.wikipedia.org/wiki/Kairos

agree it is unlike LLMs but I think it is a very interesting comparison to make, definitely made me want to chime in hahah

@ben.sless It's a thought experiment, not a white paper. It's sometimes useful to examine the logical ends to understand the middle.

Thought experiments pump intuitions. I agree that the astrology analogy appears inflammatory. I also agree that it is somewhat unfair (though my intuitions agree with the sentiment). Using astrology can be a subjective/productive activity. (Though the same thing can be and is said about using LLMs in a productive way!)

I think astrology requires engagement in a given practice to produce a given artifact, where LLMs do not: they can just produce the thing with minimal engagement. And their whole value proposition (for their owners) is the elimination of the required engagement.

@ben.sless the analogy exists in part to probe how (and how much) one trusts experts as validators instead of doers. For that purpose, strained (or even inflammatory) analogies are warranted.

There's another issue here which over strains the analogy in my opinion, which is intended use and target audience. This is the book equivalent of Personal Software. The author shared it "as is with no warranty". It is not meant to instruct or take up the collective thought space on the subject. So he is right to be surprised at the vehemence of the objections to what amounts to sharing a personal project.

This is why N-way Socratic dialog is so hard, umpteen different people and ideas end up lumped together. I have no objection (vehement or otherwise) to people sharing AI written books. I do think identifying them as such matters, which is why my analogy was about readers choosing whether to read a book and not about readers choosing when to object to the actions of content creators.

damn we complected the dialog

@danieleneal I think I'm still figuring out how best to use AI in a way that is helpful without being overreliant on its input. I've used it to varying degrees in different projects and it doesn't make me feel hampered in understanding the architectural decisions and it lets me focus on the aspect of the code that I care about, which I usually hand refine or table if it works but in a way that I find to be inoptimal. So no I'd say it doesn't make me hesitant to change or implement code in a different way for me; but it's hard to answer this question succinctly because it's pretty personal and changing and I'm still figuring it out myself. I consider the pedagogy about LLMs and AI a lot since they are here, and I remember going through high school and many teachers were grappling with Wikipedia. By the time I graduated, it seemed there was an unspoken consensus that it's reasonable to go to wikipedia as a first place in order to find primary texts. I was taught many times over not to be overreliant on computers or Google to tell me important things. Yet still, you ask people programming questions and 99% of the time the answer is “did you try googling it?” I don't think anyone in their right mind recommends AI as a panacea, and I personally think implementing it as an educational tool is easily misguided and harmful. I would probably never tell a beginner to use AI to learn how to code, but I would definitely tell them that it is one way you can figure out what is wrong with broken code. This leads me to my next thing which I'll keep brief but it's the same idea in response to the sources @stuarthalloway provides, which is again that those studies are for kids in secondary education and “novice workers”. I appreciate the peer reviewed studies proving the detrimental effects on the developmental mind; but I still anecdotally reject cognitive-surrender implicit in using LLMs; although I do still think it's a concern for myself and anyone else. Yet we come to Socrates (one of my heroes) and my favorite teaching of his which is that you cannot choose which food is served to you in this life, you can only choose whether or not to abstain from it. This is why children should have their diet moderated by an adult until such a time that the child can tend to their own health (or mind-garden, in the case of AI). I do agree though, that the n-many discussion did veer that off of the original claim which is simply that of the 3 categories of texts you described that most rational people would rely only on category one; and I still disagree but if I'm being real, I like reading so I'd probably read all 3 books just to be supportive of the non-expert, and I'd try to help them write a better book. I wanted to share the passage I was thinking of: If you had to entrust your body to someone, taking the risk of its being made better or worse, you would first consider most carefully whether you ought to entrust it or not, and would seek the advice of your friends and relations and ponder it for a number of days: but in the case of your soul, which you value much more highly than your body, and on which depends the good or ill condition of all your affairs, according as it is made better or worse, would you omit to consult first with either your father [313b] or your brother or one of us your comrades,—as to whether or no you should entrust your very soul to this newly-arrived foreigner; but choose rather, having heard of him in the evening, as you say, and coming to me at dawn, to make no mention of this question, and take no counsel upon it—whether you ought to entrust yourself to him or not; and are ready to spend your own substance and that of your friends, in the settled conviction that at all costs you must converse with https://topostext.org/people/593, whom you neither know, as you tell me, nor have ever met in argument before…

@wistfulmelodylispfulp oh that is interesting that you seem a bit immune to the thing that affects me. Do you do anything in particular to maintain the ability to go in and edit llm code? I get such a strong aversion to it in a way I never do from eg stack overflow or a blog post. Maybe a matter of time ? Did this get easier for you? Even for a simple static site I get it, eg with the CSS re: wikipedia, interesting comparison. I was definitely taught at university that using wikipedia was against the rules ... and then got a bit upset after spending hours in the library, that folk who used wikipedia got better marks. I can only begin to imagine the chaos in higher education right now...

@danieleneal I'll dm you since the thread is already long. But the tl;dr for everybody else is “I try my darndest!”

Would producing the cheat-sheet yourself have: helped you internalize the concepts more sturdily? been feasible given other demands on your time? Does the cheat-sheet list: leave out anything essential? How can you tell?

Are we talking about expertise used to write new books*? And am I an individual or a company? As an individual, I would lean heavily to developing more expertise. I would also be tempted to toy with astrology to see if there’s anything in my stars that’s worthwhile, but I wouldn’t put much weight behind it. (Why? Curiosity, I guess) If the crystal ball showed trends of using astrology and the expert writer, or heaven forbid, astrology and the non-expert writer, I would still develop expertise. But I would have self-doubt. If I were new, I might not have the judgement or confidence to choose expertise. And if I had to use astrology in order to eat, I probably would. To extend this metaphor a bit, if pilots were flying a plane and every once in a while used astrology to decide what to do, I’d be very concerned, even if the stars were mostly correct. I’d be concerned no matter what the crystal ball told me was trending. (* vs expertise gained from reading a book)

The most important trend is the one that determines: How will I acquire the food and shelter I and my loved ones require? If the broader trend is toward devaluing expertise (and instrumentalizing the time that you could have previously spent acquiring it) in favor of producing high quantities of low-value extrusions, and I must do that to pay the bills, then that is likely the way I would trend. Needs must. If you have the privilege of ignoring social trends, then you definitely should do so and do things that make you an expert, and perhaps help others get into a similar position.

The section titled "Know your leverage and use caution if you're asked back" is thoughtful. (Oh, to have a fragment link!) I think there's an opportunity here for software developers to push back and demand (!?) being treated (collectively?) as humans and not disposable human resources. Can we hang together this time?

It can be great opportunity for sure. But I don't think we have enough empathy to each other to make it happen 😐

and I think we need to outlive bubble burst first before thinking about job market flip. anyway, this article may decrease level of anxiety a little bit as it did to me, this is the main reason I decided to share it here 🙂

Timing is everything. I'm not sure I'm at the right point in my life where I can survive this particular bubble. (Shades of the market can remain irrational longer than you can remain solvent.) I truly am glad that you find the article decreases your anxiety, though. We can all use that.

@grzm FWIW, at least in Chrome and in Firefox you can select the relevant text and choose "copy link to highlight" in the context menu, to get something like this: https://apnews.com/article/ai-jobs-layoffs-hiring-rebounding-skills-04662acffb7beba822129118dde3f0a7#:~:text=Know%20your%20leverage%20and%20use%20caution%20if%20you%E2%80%99re%20asked%20back

Both. As well as how the economy as a whole is tied to so much of this. Software may be closer to the front, but few are actually unaffected by this whole thing.

@grzm I know of which you speak (myself)

I also know a new grad that has taken the stand that they wouldn't use agentic coding (but would use a chatbot), at least until they gained expertise. Still looking for a job. They're thinking of going into the trades. It's a tough market to be picky in. :(

@U0BGC50RM26 thanks for sharing! definitely worth following. but it's not that easy to join in my case, as once I could back to home, quoting Karl Marx may cost me up to 5 years in prison 😁 Yesterday I bought a tech book where author stated using llm is not a good idea. This is the level of activism I can afford atm, but keep looking for a way to go further 🙂

@grzm > but few are actually unaffected you mean yet?

The amount of money and investment in the market surrounding LLMs, the data center buildout and other directly involved projects has ramifications for the rest of the market. Capital available for other enterprises, pension funds, index funds, real estate, energy—the list goes on. Similarly, the 2008 financial crises didn't affect only the housing market.

I understand the article, but honestly, it reads as a dystopian nightmare. > Figure out what gives you a competitive edge as a human I don't even know how to explain how weird this phrase is...

@mauricio.szabo Yeah. Dystopian. I think people realize that we need what Galbraith called https://en.wikipedia.org/wiki/Countervailing_power right now, but they don't know what* that is. We're told that there is no alternative (ht Doctorow) to the path that bosses have chosen so we're all trying to scramble around trying not to get fired. I can imagine an alternative world where Ford had QA engineers using AI to help them at their job. Now Ford could make even better quality cars and sell more of them. I think everyone would be happier. (* Cue @U0BGC50RM26’s unionization, which I also support, but unions don't have much leverage anymore in the US.)

The "no alternative" bothers me way more than everything else. There's "no alternative" for a path that literally didn't exist two years ago; "inevitability" of something we can just decide to not use. But climate change? Death itself? These people don't believe these are inevitable, somehow...