This stuff is still in its infancy, of course its not perfect

But its already USEFUL and it CAN do a lot of stuff - just not all types of stuff and it still can mess up the stuff that it can do

It's that simple

The point is that overtime it'll get better and better

Reminds me of self driving cars and or even just general automation back in the day - the complaint has always been that a human could do it better and at some point those people just went away because it stopped being true

Another example is automated mail sorting by the post office. The gripe was always humans will always be able to do it better - true, in the meantime the post office reduced the facilities with humans that did this to just one

Stockfish NNUE was announced to be 80 ELO higher than the default. I don't find it frustrating. NNs excel at detecting patterns in a well defined search space.

Writing evaluation functions is tedious. It isn't a sign of NN intelligence.

I built Brokk to maximize the ability of humans to effectively supervise their AI minions. Not a VS code plugin, we need something new. https://brokk.ai

There is a class of developers who are blindly dumping the output of LLMs into PRs without paying any attention to what they are doing, let alone review the changes. This is contributing to introducing accidental complexity in the form of bolting on convoluted solutions to simple problems and even introducing types in the domain model that make absolutely no sense to anyone who has a passing understanding of the problem domain. Of course they introduce regressions no one would ever do if they wrote things by hand and tested what they wrote.

I know this, because I work with them. It's awful.

These vibecoders force the rest of us to waste eve more time reviewing their PRs. They are huge PRs that touch half the project for even the smallest change, they build and pass automated tests, but they enshitify everything. In fact, the same LLMs used by these vibecoders start to struggle how to handle the project after these PRs are sneaked in.

It's tiring and frustrating.

I apologize for venting. It's just that in this past week I lost count of the number of times I had these vibecoders justifying shit changes going into their PRs as "but Copilot did this change", as if that makes them any good. I mean, a PR to refactor the interface of a service also sneaks in changes to the connection string, and they just push the change?

In the end, I had to go spend a couple hours reading the documentation to understand the matching engine, and the final patch didn't look anything at all like the LLM generated code. Which is what seems to happen all the time, it is wonderful for spewing the boilerplate, the actual problem solving portions its like talking to someone who simply doesn't understand the problem and keeps giving you what it has, rather than what you want.

OTOH, its fantastic for review/etc even though I tend to ignore many of the suggestions. Its like a grammar checker of old, it will point out you need a comma you missed, but half the time the suggestions are wrong.

- LLMs are going to be 100x more valuable than me and make me useless? I don't see it happening. Here's 3 ways I'm still better than them.

Another factor is the capture of market sectors by Big Co. When buyers can only approach some for their products/services, the Big Co can drastically reduce quality and enshittify without hurting the bottom line much. This was the big revelation when Elon gutted Twitter.

And so we are in for interesting times. On the plus side, it is easier than ever to create software and distribute it. Hiring doesn't matter if I can get some product sense and make some shit worth buying.

You know, those who don't care about learning and solving problems, gaining real experience they can use to solve problems even faster in the future, faster than any AI slop.

LLM is as good as the material it is being trained on, the same applies to AI and they are not perfect.

Perplexity AI did assist me in getting into Python from 0 to getting my code test with 94% covered and no vulnerabilities (scanning tools) Google Gemini is dogshit

Trusting blindly into a code generated by LLM/AI is a whole complete beast, and I am seeing developers doing basically copy/paste into company's code. People are using these sources as the truth and not as a complementary tool to improve their productivity.

I memorize really little and tend to spend time on reinventing algorithms or looking them up in documentation. Verifying is easy except the fee cases where the llm produces something really weird. But then fallback to docs or reinventing.

A chunk of the objections indicate people trying to shoehorn in their old way of thinking and working.

I think you have to experiment and develop some new approaches to remove the friction and get the benefit.

It’s a tough bar if LLMs have to be post antirez level intelligence :)

Glad to see the author acknowledges their usefulness and limitations so far.

At this point I think people who don't see the value in AI are willfully pulling the wool over their own eyes.

Think about it and tell me you use it differently.

General LLM model would not be as good as LLM for coding, for this case Google deepmind team maybe has something better than Gemini 2.5 pro

In that regard I am less optimistic.

We have never been good at confronting the follies of management. The Leetcode interview process is idiotic but we go along with it. Ironically LC was one of the first victims of AI, but this is even more of an issue for management that things SWEs solve Leetcodes all day.

Ultimately I believe this is something that will take a cycle for business to figure out by failing. When businesses will figure out that 10 good engineers + AI always beats 5 + AI, it will become table stakes rather than something that replaces people.

Your competitor who didn’t just fire a ton of SWEs? Turns out they can pay for Cursor subscriptions too, and now they are moving faster than you.

In my experience working on enterprise app modernization, we’ve found success by keeping humans firmly in the loop. Tools like Project Analyzer (from Techolution’s AppMod.AI suite) assist engineers in identifying risky legacy code, mapping dependencies, and prioritizing refactors. But the judgment calls, architecture tweaks, and creative problem-solving? Still very much a human job.

LLMs boost productivity, but it’s the developer's thinking that makes the outcome truly resilient. This story captures that balance perfectly.

Increase the temperature of the LLMs.

Ask several LLMs, each several time the same question, with tiny variations. Then collect all answers, and do a second/third round asking each LLM to review all collected answers and improve.

Add random constraints, one constraints per question. For example, to LLM: can you do this with 1 bit per X. Do this in O(n). Do this using linked lists only. Do this with only 1k memory. Do this while splitting the task to 1000 parallel threads, etc.

This usually kicks the LLM out of its confort zone, into creative solutions.

How did it help, really? By telling you your idea was no good?

A less confident person might have given up because of the feedback.

I just can't understand why people are so excited about having an algorithm guessing for them. Is it the thrill when it finally gets something right?

If software is about meeting human demands, humans will always write its requirements, by definition. If we build another machine like LLMs, well the design of those LLMs is subject to human demands. There is no point at which we can demand perfection but not be involved in its definition.

LLMs are great as assistants. Just today, Copilot told me it's there to do the "tedious and repetitive" parts so I can focus my energy on the "interesting" parts. That's great. They do the things every programmer hates having to do. I'm more productive in the best possible way.

But ask it to do too much and it'll return error-ridden garbage filled with hallucinations, or just never finish the task. The economic case for further gains has diminished greatly while the cost of those gains rises.

Automation killed tons of manufacturing jobs, and we're seeing something similar in programming, but keep in mind that the number of people still working in manufacturing is 60% of the peak, and those jobs are much better than the ones in the 1960s and 1970s.

Translation software has been around for a couple of decades. It was pretty shitty. But about 10 years ago it started to get to the point where it could translate relatively accurately. However, it couldn't produce text that sounded like it was written by a human. A good translator (and there are plenty of bad ones) could easy outperform a machine. Their jobs were "safe".

I speak several languages quite well and used to do freelance translation work. I noticed that as the software got better, you'd start to see companies who instead of paying you to translate wanted to pay you less to "edit" or "proofread" a document pre-translated by machine. I never accepted such work because sometimes it took almost as much work as translating it from scratch, and secondly, I didn't want to do work where the focus wasn't on quality. But I saw the software steadily improving, and this was before ChatGPT, and I realized the writing was on the wall. So I decided not to become dependent on that for an income stream, and moved away from it.

When LLMs came out, and they now produce text that sounded like it was written by a native speaker (in major languages). Sure, it's not going to win any literary awards, but the vast vast majority of translation work out there is commercial, not literature.

Several things have happened: 1) there's very little translation work available compared to before, because now you can pay only a few people to double-check machine-generated translations (that are fairly good to start with); 2) many companies aren't using humans at all as the translations are "good enough" and a few mistakes won't matter that much; 3) the work that is available is high-volume and uninteresting, no longer a creative challenge (which is why I did it in the first place); 4) downward pressure on translation rates (which are typically per word), and 5) very talented translators (who are more like writers/artists) are still in demand for literary works or highly creative work (i.e., major marketing campaign), so the top 1% translators still have their jobs. Also more niche language pairs for which LLMs aren't trained will be safe.

It will continue to exist as a profession, but diminishing, until it'll eventually be a fraction of what it was 10 or 15 years ago.

(This is specifically translating written documents, not live interpreting which isn't affected by this trend, or at least not much.)

So much of it is exploratory, deciding how to solve a problem from a high level, in an understandable way that actually helps the person who it’s intended to help and fits within their constraints. Will an LLM one day be able to do all of that? And how much will it cost to compute? These are the questions we don’t know the answer to yet.

Like sex professionals, Gemini and co are made to be impressed and have possitive things to say about programming ideas you propose and find your questions "interesting", "deep", "great" and so.

There is a principle (I forget where I encountered it) that it is not code itself that is valuable, but the knowledge of a specific domain that an engineering team develops as they tackle a project. So code itself is a liability, but the domain knowledge is what is valuable. This makes sense to me and matched my long experience with software projects.

So, if we are entrusting coding to LLMs, how will that value develop? And if we want to use LLMs but at the same time develop the domain acumen, that means we would have to architects things and hand them over to LLMs to implement, thoroughly check what they produce, and generally guide them carefully. In that case they are not saving much time.

Doing the actual thinking is generally not the part I need too much help with. Though it can replace googling info in domains I'm less familiar with. The thing is, I don't trust the results as much and end up needing to verify it anyways. If anything AI has made this harder, since I feel searching the web for authoritative, expert information has become harder as of late.

LLMs are faster, and when the task can be synthetically tested for correctness, and you can build up to it heuristically, humans can't compete. I can't spit out a full game in 5 minutes, can you?

LLMs are also cheaper.

LLMs are also obedient and don't get sick, and don't sleep.

Humans are still better by other criteria. But none of this matters. All disruptions start from the low end, and climb from there. The climbing is rapid and unstoppable.

Maybe LLMs completely trivialize all coding. The potential for this is there

Maybe progress slows to a snails pace, the VC money runs out and companies massively raise prices making it not worth it to use

No one knows. Just sit back and enjoy the ride. Maybe save some money just in case

The other, related question is, are human coders with an LLM better than human coders without an LLM, and by how much?

(habnds made the same point, just before I did.)

There will always be a place for really good devs but for average people (most of us are average) I think there will be less and less of a place.

While technically capable of building it on my own, development is not my day job and there are enough dumb parts of the problem my p(success) hand-writing it would have been abysmal.

With rose-tinted glasses on, maybe LLM's exponentially expand the amount of software written and the net societal benefit of technology.

Maybe the way forward would be to invent better "specifiction languages" that are easy enough for humans to use, then let the AI implement the specifciation you come up with.

Also, so many people said the same thing about chess when the first chess programs came out. "It will never beat an international master." Then, "it will never beat a grandmaster." And Kasparov said, "it would never beat me or Karpov."

Look where we are today. Can humanity adapt? Yes, probably. But that new world IMO is worse than it is today, rather lacking in dignity I'd say.

I realize how lucky I am to even have a job that I thoroughly enjoy, do well, and get paid well for. So I'm not going to say "It's not fair!", but ... I'm bummed.

As a software engineer, I need to solve business problems, and much of this requires code changes, testing, deployments, all that stuff we all know. Again, if a good AI could take on a lot of that work, maybe that means I don't have to sit there in dependency hell and fight arcane missing symbol errors for the rest of my fucking career.

I still think about Tom Scott's 'where are we on the AI curve' video from a few years back. https://www.youtube.com/watch?v=jPhJbKBuNnA

Looking forward for rubber duck shaped hardware AI interfaces to talk to in the future. Im sure somebody will create it

It's ignorant to think machines will not catch up to our intelligence at some point, but for now, it's clearly not.

I think there needs to be some kind of revolutionary breakthrough again to reach the next stage.

If I were to guess, it needs to be in the learning/back propagation stage. LLM's are very rigid, and once they go wrong, you can't really get them out of it. A junior develop for example could gain a new insight. LLM's, not so much.

They drive you nuts trying to communicate with them what you actually want them to do. They have a vast array of facts at immediate recall. They’ll err in their need to produce and please. They do the dumbest things sometimes. And surprise you at other times. You’ll throw vast amounts of their work away or have to fix it. They’re (relatively) cheap. So as an army of monkeys, if you keep herding them, you can get some code that actually tells a story. Mostly.

"UnitedHealth uses AI model with 90% error rate to deny care, lawsuit alleges" Also "The use of faulty AI is not new for the health care industry."

It would actually have been more pernicious that way, since it would lull people into a false sense of security.

Yep - my number 1 use case for LLMs is as a template and example generator. It actually seems like a fairly reasonable use for probabilistic text generation!

> At least with humans I have some basic level of trust, so that even if I don't understand the code at that level, I can scan it and see that it's reasonable.

If you can't scan the code and see that it's reasonable, that's a smell. The task was too big or its implemented the wrong way. You'd feel bad telling a real person to go back and rewrite it a different way but the LLM has no ego to bruise.

I may have a different perspective because I already do a lot of review, but I think using LLMs means you have to do more of it. What's the excuse for merging code that is "off" in any way? The LLM did it? It takes a short time to review your code, give your feedback to the LLM and put up something actually production ready.

> But every piece of LLM generated code I've seen to date hasn't been trustworthy once I put in the effort to really understand it.

That's why your code needs tests. More tests. If you can't test it, it's wrong and needs to be rewritten.

My approach is to describe the task in great detail, which also helps me completing my own understanding of the problem, in case I hadn't considered an edge case or how to handle something specific. The more you do that the closer the result you get is to your own personal taste, experience and design.

Of course you're trading writing code vs writing a prompt but it's common to make architectural docs before making a sizeable feature, now you can feed that to the LLM instead of just having it be there.

I think we are, especially with executives mandating the use LLMs use and expecting it to massively reduce costs and increase output.

For the most part they don't actually seem to care that much about software quality, and tend to push to decrease quality at every opportunity.

Yeah we shouldn’t and I limit my usage to stuff that is easily verifiable.

But there’s no guardrails on this stuff, and one thing that’s not well considered is how these things which make us more powerful and productive can be destructive in the hands of well intentioned people.

I'm betting they're the most senior people on the team.

I have been drawing all my life and studied traditional animation though, so it’s probably a little bit of nature and nurture.

I genuinely think this could be great for toys that kids grow up with i.e. the toy could adjust the way it talks depending on the kids age and remember key moments in their life - could be pretty magical for a kid

> you will almost certainly be replaced by an llm in the next few years

So... Maybe not. I agree that Hacker News does have a generally higher quality of contributors than many places on the internet, but it absolutely is not a universal for HNers. There are still quite a few posters here that have really bought into the hype for whatever reason

But he should've know that people would jump at the opportunity to contradict him and should've written his comment so as not to admit such an easily-contradictable interpretation.

In fact, maybe most of has have been replaced by LLMs already :-)

Personally, I'm more interested in the political angle. I can see that AI will be disruptive because there's a ton of money and possibly other political outcomes depending on it doing exactly that.

> it was a decent, if probabilistic, rubber duck

How is it a rubber duck if it suggested where the problem was?

Isn't a rubber duck a mute object which you explain things to, and in the process you yourself figure out what the solution is?

That was the way I saw it for a while. In recent months I've begun to wonder if I need to reevaluate that, because it's become clear to me that scaling doesn't actually start from zero. By zero I mean that I was naive enough to think that all programs, even the most googled programmed one by a completely new junior would at least have, some, efficiency... but some of these LLM services I get to work on today are so bad they didn't start at zero but at some negative number. It would have been less of an issue if our non-developer-developers didn't use Python (or at least used Python with ruff/pyrefly/whateveryoulike, but some of the things they write can't even scale to do minimal BI reporting.

It eludes all of those who died in the process : those still alives are here despite bad IT, not due to a bad IT

Recently I converted all the (Google Docs) documentation of a project to markdown files and added those to the workspace. It now indexes it with RAG and can easily find relevant bits of documentation, especially in agent mode.

It really stresses the importance of getting your documentation and processes in order as well as making sure the tasks at hand are well-specified. It soon might be the main thing that requires human input or action.

Any skilled developer with a decade of experience can write prompts that return back precisely what we wanted almost every single time. I do it all day long. "Claude 4" rarely messes up.

Poker tests intelligence. So what gives? One interesting thing is that for whatever reason, poker performance isn't used a benchmark in the LLM showdown between big tech companies.

The models have definitely improved in the past few years. I'm skeptical that there's been a "break-through", and I'm growing more skeptical of the exponential growth theory. It looks to me like the big tech companies are just throwing huge compute and engineering budgets at the existing transformer tech, to improve benchmarks one by one.

I'm sure if Google allocated 10 engineers a dozen million dollars to improve Gemini's poker performance, it would increase. The idea before AGI and the exponential growth hypothesis is that you don't have to do that because the AI gets smarter in a general sense all on it's own.

This has been the biggest boost for me. The number of choices available when facing a blank page is staggering. Even a bad/wrong implementation helps collapse those possibilities into a countable few that take far less time to think about.

We did. Why do you think labor laws, unions, etc. exist? Why do you think communism was appealing as an idea in the beginning to many? Whether the effects were good or bad or enough or not, that’s a different question. But these changes have demonstrably, grave consequences.

They don't

> Garbage in = garbage out generally.

Generally, this statement is false

> When attention is managed and a problem is well defined and necessary materials are available to it, they can perform rather well.

Keyword: can.

They can also not perform really well despite all the management and materials.

They can also work really well with loosey-goosey approach.

The reason is that they are non-deterministic systems whose performance is affected more by compute availability than by your unscientific random attempts at reverse engineering their behavior https://dmitriid.com/prompting-llms-is-not-engineering

People are expecting perfection from bad spec

Isn’t that what engineers are (rightfully) always complaining about to BD?

I've definitely also found that the poor code can sometimes be a nice starting place. One thing I think it does for me is make me fix it up until it's actually good, instead of write the first thing that comes to mind and declare it good enough (after all my poorly written first draft is of course perfect). In contrast to the usual view of AI assisted coding, I think this style of programming for tedious tasks makes me "less productive" (I take longer) but produces better code.

Not really, not always. To anyone who’s used the latest LLMs extensively, it’s clear that this is not something you can reliably assume even with the constraints you mentioned.

No they don't, they generate a statistically plausible text response given a sequence of tokens.

That's because other people are making those working well. It's like how you don't care about how the bread is being made because you trust your baker (or the regulations). It's a chain of trust that is easily broken when LLMs are brought in.

^ This. This is where I've landed as far as the extent of LLM coding assistants for me.

This "vibe coding" seems just another way to say that people spend more time refining the output of these tools over and over again that what they would normally code.

Wrote me:

- a SPI deserializer that sets a bit after 12 bits read in, to trigger a prefetch

- an SDC constraints file for the deserializer that correctly identified the SPI clock and bus clock as separate domains requiring their own statement

- a test bench that validated both that the prefetch bit was being set, and that it was being set at the proper time relative to the SPI clock

- a makefile with commands for build, headless test, and debug by loading the VCD into a waveform viewer

It always feels like LLMs can do lots of the easy stuff, but if they can't do everything you need the skilled engineer anyway, who'd knock the easy things out in a week anyway.

Nearly every part of the tool flow I just described, I would consider “tricky to get right”. Been doing this for ~15 years and it’s still tough to bootstrap something like this from scratch. ChatGPT-4o did this for me from zero in about 15 minutes.

I won’t lie: I love it. I can focus on the actual, bigger problems at hand, and not the tricky little details of HDLs.

People are either deluding themselves or ignorant of the capabilities of frontier models if they don’t believe LLMs offer a speedup in workflow.

I personally believe that most of the doubt and cynicism is due to:

1) a pretty big collective identity crisis among software professionals, and

2) a suspicion that LLMs make it so that anyone who is good at articulating the problem precisely no longer needs a software engineer as a translation specialist from specs to code.

I say this as an EE of ~15 years who’s always been able to articulate what I want, specifically, to a firmware counterpart, who then writes the code I need. I can turn years of practice in this skill into great prompts for an LLM, which effectively cuts out the middleman.

I really like it. It’s helped me take on a lot of projects that are just outside of my innate level of capability. It’s also helped me learn a lot of new things about these new software adjacent areas. ChatGPT is a great tutor!

In this example the idea that losing the muscles that make you great at hoeing" seems kind of like a silly thing to worry about

But I think there's a second order effect here. The kid gets a job driving the tractor instead. He spends his days seated instead of working. His lifestyle is more sedentary. He works just as many hours as before, and he makes about the same as he did before, so he doesn't really see much benefit from the increased productivity of the tractor.

However now he's gaining weight from being more sedentary, losing muscle from not moving his body, developing lower back problems from being seated all day, developing hearing loss from the noisy machinery. His quality of life is now lower, right?

Edit: Yes, there are also health problems from working hard moving dirt all day. You can overwork yourself, no question. It's hard on your body, being in the sun all day is bad for you.

I would argue it's still objectively a physically healthier lifestyle than driving a tractor for hours though.

Edit 2: my point is that I think after driving a tractor for a while, the kid would really struggle to go hoe by hand like he used to, if he ever needed to

I think it is about how utilitarian the output is. For food no one cares how the sausage is made. For a painting the story behind it is more important than the picture itself. All of Picasso's paintings are famous because they were painted by Picasso. Picasso style painting by Bill? Suddenly it isn't museum worthy anymore.

No one cares about the story or people behind Word, they just want to edit documents. The Demo scene probably has a good shot at being on the side of art.

You can let it self-drive, but you'd probably learn nothing, and it will actually take you longer. Put an expert driver behind the wheel, and they'll drive faster and only use automation features for the boring parts.

Which means it's back to being a very useful tool, but not the earth-shattering disruptor we hoped (or worried) it would be.

Clearly they do work in a general sense. People who don't want to code are making things that work this way right now!

This isn't yet replacing me, but I'm certain it will relatively soon be the standard for how software is developed.

Using Google to find an answer is convenient, and I'm sure you would miss it.

But telling a machine to think for you outsources everything, once it's gone you have nothing left.

Because you don't have a choice. Your thoughts are not code.

"AI" is the latest iteration of snake oil that is foisted upon us by management. The problem is not "AI" per se, but the amount of of friction and productivity loss that comes with it.

Most of the productivity loss comes from being forced to engage with it and push back against that nonsense. One has to learn the hype language, debunk it, etc.

Why do you think IT has gotten better? Amazon had a better and faster website with far better search and products 20 years ago. No amount of "AI" will fix that.

Depends on the context. You have to keep in mind: it is not a goal of our society or economic system to provide you with a stable, rewarding job. In fact, the incentives are to take that away from you ASAP.

Before software engineers go celebrate this tech, they need realize they're going to end up like rust-belt factory workers the day after the plant closed. They're not special, and society won't be any kinder to them.

> ...and even wiser to be in charge of and operate the replacement.

You'll likely only get to do that if your boss doesn't know about it.

For other stuff this is obviously harder.

But these endless claims that the fact they're "just" predicting tokens means something about their computational power are based on flawed assumptions.

So the argument goes: LLMs were trained to predict the next token, and the most general solution to do this successfully is by encoding real understanding of the semantics.

It's even crazier that some people believe that humans "evolved" intelligence just by nature selecting the genes which were best at propagating.

Clearly, human intelligence is the product of a higher being designing it.

/s

Statistical text (token) generation made from an unknown (to the user) training data set is not the same as a keyword/faceted search of arbitrary content acquired from web crawlers.

> The problem is that the conversational interface, for some reason, seems to turn off the natural skepticism that people have when they use a search engine.

For me, my skepticism of using a statistical text generation algorithm as if it were a search engine is because a statistical text generation algorithm is not a search engine.

Recently I did some tests with coding agents, and being able to translate a full application from AT&T Assembly into Intel Assembly compatible with NASM, in about half an hour of talking with agent, and having the end result actually working with minor tweeks isn't something a "decent search engine a la Google circa 2005." would ever been able to achieve.

In the past I would have given such a task to a junior dev or intern, to keep them busy somehow, with a bit more tool maturity I have no reason to do it in the future.

And this is the point many developers haven't yet grasped about their future in the job market.

  > the conversational interface, for some reason, seems to turn off the natural skepticism that people have

No, the top players when the meta changes in competitive games remain the top players. They also figure out the new meta faster than the casual players.

People with a minimum amount of expertise stop asking for advice for average circumstances very quickly.

In a healthy competitive market (like most of the history of the US, maybe not the last 30-40 years), if all of the farms do that, it drives down the cost of the food. The reduction in labor necessary to produce the food causes competition and brings down the cost to produce the food.

That still doesn’t directly benefit the farmhands. But if it happens gradually throughout the entire economy, it creates abundance that benefits everybody. The farmhand doesn’t benefit from their own increase in productivity, but they benefit from everyone else’s.

And those unemployed farmhands likely don’t stay unemployed - maybe farms are able to expand and grow more, now that there is more labor available. Maybe they even go into food processing. It’s not obvious at the time, though.

In tech, we currently have like 6-10 mega companies, and a bunch of little ones. I think creating an environment that allows many more medium-sized companies and allowing them to compete heavily will ease away any risk of job loss. Same applies to a bunch of fields other than tech. The US companies are far too consolidated.

Sure, although I think our lives are generally better than they were a few hundred years ago. Besides, if you care about your health you can always take steps yourself.

> The only one who benefits are the owners

Well yeah, the entity that benefits is the farm, and whoever owns whatever portions of the farm. The point of the farm isn't to give its workers jobs. It's to produce something to sell.

As long as we're in a market where we're selling our labor, we're only given money for being productive. If technology makes us redundant, then we find new jobs. Same as it ever was.

Think about it: why should hundreds of manual farmhands stay employed while they can be replaced by a single machine? That's not an efficient economy or society. Let those people re-skill and be useful in other roles.

An LLM is more like outsourcing to a consultancy. Results may vary.

Tools and systems which increase productivity famously always put everyone out of a job, which is why after a couple centuries of industrial revolution we're all unemployed.

When talking with reasonable people, they will tell you if they don't understand what you're saying.

When talking with reasonable people, they will tell you if they don't know the answer or if they are unsure about their answer.

LLMs do none of that.

They will very happily, and very confidently, spout complete bullshit at you.

It is essentially a lotto draw as to whether the answer is hallucinated, completely wrong, subtly wrong, not ideal, sort of right or correct.

An LLM is a bit like those spin the wheel game shows on TV really.

I could also argue that the word "design" has a connotation strictly opposing emergent behaviour like evolution, as in the intelligent design "theory". So not the best word to use perhaps.

And in your example, just because we made a system that exhibits emergent behaviour to some degree, we can't assume it can "design" intelligence the way evolution did, on a much, much shorter timeline, no less.

I do concur it is an individualistic strategy, and as you mentioned unionization might have helped. But, then again it might not. Developers are partially unionized where I live, and I'm not so sure it's going to help. It might absorb some of the impact. Let's see in a couple of years.

Google search includes an AI generated response.

Gemini prompts return Google search results.

But I actually can’t imagine how you can teach someone to code if they have access to an LLM from day one. It’s too easy to take the easy route and you lose the critical thinking and problem solving skills required to code in the first place and to actually make an LLM useful in the second. Best of luck to you… it’s a weird time for a lot of things.

*edit them/they

TBF, trial and error has usually been my path as well, it's just that I was generating the errors so I would know where to find them.

It's hard to remember what it was like to be in that phase. Once simple things like using variables are second nature, it's difficult to put yourself back into the shoes of someone who doesn't understand the use of a variable yet.

I do use these tools though! I spent some time with AI. I have coworkers who are more heads-down working on their projects and not tinkering with agents, and they're doing fine. I have coworkers who are on the absolute bleeding edge of AI tools, and they're doing fine. When the tooling matures and the churn lessens and the temperature of the discourse is lowered, I'm confident that we will all be doing great things. I just think that the "anybody not using and optimizing Codex or Claude Code today is not gonna make it" attitude is misguided. I could probably wring out some more utility from these tools if I spent more time with them, but I'd rather spend most of my professional development time working on subject matter expertise. I want to deeply understand my domain, and I trust that AI use will (mostly) become relatively easier to pick up and less of a differentiator as time goes on

I think parent is agreeing with you?

> This is why devs who started with J2ME are the holy grail of app developers, since they started making apps years before iPhone devs

Look, we see the forest. We are just not impressed by it.

Having unlimited chaos monkeys at will is not revolutionizing anything.

There's no guarantee a technology will take off, even if it's really, really good. Because we don't decide if that tech takes off - the lawyers do. And they might not care, or they might decide billing more hours is better, actually.

When I graduated high school, I had never been or knew anyone who had ever been on the internet at all. The internet was this vague "information superhighway" that I didn't know really what to make of.

If you are of a certain age though you would think a pointless update to react was all the change ever coming.

That time is over and we are back to reality.

Or maybe they just know the nitty-gritty inherent limitations of technology better than you.

(inb4: "LLMs can't have limitations! Wait a few years and they will solve literally every possible problem!")

Consider, how it's been done traditionally for imperative programming: you explain the notion of programming (encoding algorithms with a specific set of commands),explain basic control flow, explain flowcharts, introduce variables and a simplified computation model. Then you drop the student into a simplified environment where they can test the basics in practice, without the need to use any "incantations".

By the time you need to introduce `#include <stdio.h>` they already know about types, functions, compilation, etc. At this point you're ready to cover C idioms (or any other language) and explain why they are necessary.

If ChatGPT needs to, it will actually do the search for me and then collate the results.

IMO, they're still useless today, with the only progress being that they can produce a more convincing facade of usefulness. I wouldn't call that very meaningful progress.

LLMs can only give you code that somebody has wrote before. This is inherent. This is useful for a bunch of stuff, but that bunch won't change if OpenAI decides to spend the GDP of Germany training one instead of Costa Rica.

Programming has become vastly more efficient in terms of programmer effort over decades, but making some aspects of the job more efficient just means all your effort it spent on what didn’t improve.

no i don't remember that. They are doing similar things now that they did 3 yrs ago. They were still a decent rubber duck 3 yrs ago.

This doesn't feel like that. The applications of generative AI have become self-evident to anyone that's followed their rise. Specific applications of AI resemble snake oil, and there are hucksters who pivoted from crypto to AI, but the ratio of legit use cases to scams isn't even close.

If anything, the incentives for embellishment have flipped since crypto. VC-funded AI companies will dreamily fire press releases about AI taking us to Mars, but it doesn't have the pseudo-grassroots quality of cryptocurrency hype. The average worker is incentivized to be an AI skeptic. The rise of generative AI threatens workers in several fields today, and has already negatively impacted copywriters and freelance artists. I absolutely understand why people in those fields would respond by calling AI use unethical and criticize the shortcomings of today's models.

We'll see what the next few years hold. But personally, I foresee AI integration ramping up. Even if the models themselves completely stagnate from this point on, there's a lot of missing glue between the models and the real world.

Don't people and companies using AI lazily to put out low quality content blind you to its potential as well as the reality of what it can do right now. Look at Google's VO3, most people in the world right now won't be able to tell you that it's AI generated and not real.

I don't need to convince anyone that LLMs are enabling me to do a lot more. This is what makes this hype different. It has bones. Once you've found a way to leverage it, it's undeniably helpful regardless of your prior disposition. Everyone else can say they're not useful and it rings hollow because it obviously is to me. And thus probably useful to everyone else too.

x10 of zero is still zero, I guess.

When it's time to learn Java you're supposed to be past the basics. Old-school intros to programming starts with flowcharts for a reason.

You can learn either way, of course, but with one, people get tied up to a particular language-specific model and then have all kinds of discomfort when it's time to switch.

I think the mental rewiring that goes on as you move past those primitive first steps is so comprehensive that it makes it hard to relate across that knowledge boundary. Some of the hardest things to explain are the ones that have become a second nature to us.

I didn't have any programming books or even the internet back then. It was a poke and prod at the magical incantations type of thing.

Then the people who congratulate the AI for helping get yelled at by the other category.

Read the article; his younger self failed to see logic needed now. Add that onion peel. No such thing as perfect clairvoyance.

Even Yann LeCun’s energy based models driving robots have the same experience problem.

Make a computer that can observe all of the past and future.

Without perfect knowledge our robots will fail to predict some composition of space time before they can adapt.

So there’s no probe we can launch that’s forever and generally able to survive with our best guess when launched.

More people need to study physical experiments and physics and not the semantic rigor of academia. No matter how many ideas we imagine there is no violating physics.

Pop culture seems to have people feeling starship Enterprise is just about to launch from dry dock.

This !

Yeah, it probably "worked for me" because they spent a gazillion hours engaging in what the LLM fanbois call "prompt engineering", but you and I would call "engaging in endless iterative hacky work-arounds until you find a prompt that works".

Unless its something extremely simple, the chances of an LLM giving you a workable answer on the first attempt is microscopic.

But how much of enterprise programming is 'get some data from a database, show it on a Web page (or gui), store some data in the database', with variants?

It makes sense that we have libraries for abstraction away some common things. But it also makes sense that we can't abstract away everything we do multiple times, because at some point it just becomes so abstract that it's easier to write it yourself than to try to configure some library. Does not mean that it's not a variant of something done before.

Your claim here is slightly different.

You're claiming that if a token isn't supported, it can't be output [1]. But we can easily disprove this by adding minimal support for all tokens, making C appear in theory. Such support addition shows up all the time in AI literature [2].

[1]: https://en.wikipedia.org/wiki/Support_(mathematics)

[2]: In some regimes, like game theoretic learning, support is baked into the solving algorithms explicitly during the learning stage. In others, like reinforcement learning, its accomplished by making the policy a function of two objectives, one an exploration objective, another an exploitation objective. That existing cross pollination already occurs between LLMs in the pre-trained unsupervised regime and LLMs in the post-training fine-tuning via forms of reinforcement learning regime should cause someone to hesitate to claim that such support addition is unreasonable if they are versed in ML literature.

Edit:

Got downvoted, so I figure maybe people don't understand. Here is the simple counterexample. Consider an evaluator that gives rewards: F("AAC") = 1, all other inputs = 0. Consider a tokenization that defines "A", "B", "C" as tokens, but a training dataset from which the letter C is excluded but the item "AAA" is present.

After training "AAA" exists in the output space of the language model, but "AAC" does not. Without support, without exploration, if you train the language model against the reinforcement learning reward model of F, you might get no ability to output "C", but with support, the sequence "AAC" can be generated and give a reward. Now actually do this. You get a new language model. Since "AAC" was rewarded, it is now a thing within the space of the LLM outputs. Yet it doesn't appear in the training dataset and there are many reward models F for which no person will ever have had to output the string "AAC" in order for the reward model to give a reward for it.

It follows that "C" can appear even though "C" does not appear in the training data.

[0]: https://mockaroo.com/

If you tell someone that multiplying a number by 2 is adding the number to itself, then if this person knows addition, you can't be surprised if it tells you that 9*2 is 18. A small leap in discovery is when the person can extract the pattern and gives you 5*3 is 5+5+5. A much bigger leap is when the person discovers exponent.

But if you take the time to explain each concept....

The rest of what you wrote here is either also true for humans or not true for machines irrespective of your definitions unless you can demonstrate that humans can exceed the Turing computable.

You can not.

LLM's will make a lot of things easier for humans, because most of the thinking the humans do have been automated into the LLM. But ultimately you run into a limit where the human has to take over.

There is no inherent reason why they need to be.

> So yeah theoretically you could generate reward models with LLMs, but they won't be any good, unless they are curated by other reward models that are ultimately curated by humans.

This reasoning is begging the question: The reasoning is true only if the conclusion is true. It's therefore a logically invalid argument.

There is no inherent reason why this needs to be the case.

> An LLM only learns through input text.

This is false. There already exist LLM which understand more than just text. Relevant search term: multi-modality.

> It doesn't have a first-person 3D experience of the world.

Again false. It is trivial to create such an experience with multi-modality. Just set up an input device which streams that.

> So it can't execute physical experiments, or even understand them.

Here you get confused again. It doesn't follow, based on perceptual modality, that someone can't do or understand experiments. Hellen Keller can be both blind, but also do an experiment.

Beyond just being confused, you also make another false claim. Current LLMs already have the capacity to run experiments and do so. Search terms: tool usage, ReAct loop, AI agents.

> It can understand the texts about it, but it can't visualize it, because it doesn't have a visual experience.

Again, false!

Multi-modal LLMs currently possess the ability to generate images.

> And ultimately our physical world is governed by physical processes. So at the fundamentals of physical reality, the LLMs lack understanding. And therefore will stay dependent on humans educating and correcting it.

Again false. The same sort of reasoning would claim that Hellen Keller couldn't read a book, but braille exists. The ability to acquire information outside an umwelt is a capability that intelligence enables.

So such algorithms can replace certain functions of humans, but they can't replace the human as a whole. And that is the same with LLMs. They save us time for repetative tasks, but they can't replace all of our functions. In the end an LLM is a comprehensive algorithm constantly updated with machine learning. It's very helpful, but it has its limits. The limit is constantly surpassed, but it will never replace a full human. To do that you need to do a whole lot more than a comprehensive machine learning algorithm. They can get very close to something that looks like a human, but there will always be something lacking. Which then again can be improved upon, but you never reach the same level.

That is why I don't worry about AI taking our jobs. They replace certain functions, which will make our job easier. I don't see myself as a coder, I see myself as a system designer. I don't mind if AIs take over (certain parts of) the coding process (once they're good enough). It will just make software development easier and faster. I don't think there will be less demand for software developers.

It will change our jobs, and we'll have to adapt to that. But that is always what happens with new technology. You have to grow along with the changes and not expect that you can keep doing the same thing for the same value. But I think that for most software developers that isn't news. In the old days people were programming in assembly, then compiled languages came and then higher level languages. Now we have LLMs, which (when they become good enough) will just be another layer of abstraction.