Counterpoint, I sold all my Bitcoin in 2011 when Mt Gox got hacked and the price plummeted 80%. Would have done it again after their 2014 hack too if I had any left.

> Bitcoin is a good example: if you bought it 15 years ago and held it, you're probably quite wealthy by now

But you just said bail the moment it's future starts to be questionable. If you follow that you would have never held it for 15 years.

What would have you done when the Bitcoin fork happened 50/50? Would you have gone int ICOs? Which ones? Etc…

There’s simply too many “new things”, so by trying to get exposure to them you’ll be massively in the red.

Let’s say you get into 1000 “new things”, and you strike it lucky and hit BTC. You’d had to buy BTC in early 2013, hold it over the whole period and sold at the historical maximum for you to be at break even.

If instead of buying 1000 “new things”, you’ve put your money into the S&P you’d be at +250% by the same time.

For me though, I'm dabbling in AI because it fascinates me. Bitcoin was like, I don't know, Herbalife? —never interesting to me at all.

I'm still stuck with TFS and SVN in my day jobs but use Git on and off on side projects. I really wish all my clients would just switch to Git.

It's easy to say "well of course I would have invested in Google in 1999" but there was nothing in 1999 to say that Google was going to be as big as it was. Why not Lycos or Dogpile or AskJeeves?

How many people dedicated their careers to Flash, only to have it die at the hands of Steve Jobs and HTML5? It's not just about bailing out: lots of folks had to start over because taking advantage of the opportunity means actually investing real time and money. "As a tulip bulb producer, I would have simply stopped producing tulip bulbs when it started to seem questionable." https://en.wikipedia.org/wiki/Tulip_mania

The problem is this leaves you undifferentiated from every hype chaser in Silicon Valley. Our world is littered with folks who went to coding school, traded Bitcoin, did something in the metaverse and blogged about AI. That jack-of-all-trades knowledge can be useful. But only if you’re making unlikely connections. Having the same cutting-edge familiarity as every tech journalist doesn’t that make.

Better: develop deep knowledge and expertise in something. Anything. Not only does this give you some ability to recognize what expertise looks like from afar, it also lets you dip into new topics and have a chance at seeing something everyone else hasn’t already. That, in turn, gives you the ability to be a meaningful first mover.

If you sold the farm to get in early in the Metaverse, you're totally hosed now because that was a dead end. The idea of digital real estate was as terrible then as it is now.

It just sounds like a giant scheme to burn through tokens and give money to the AI corps, and tech directors are falling for it immediately.

MS Access and so many more "you won't need a programmer again" dev tools over the decades blazed the trail.

I'm still working in tech, and likely will forever in a much reduced capacity. But pottery is my life now.

That is why I agree with the sentiment as well. I use AI a little. Not too much. And I'm as swamped with work as ever because my focus is on legacy stacks, where AI is really not strong.

Meanwhile, the main category of people who have consistently gotten rich off the "crypto revolution" were various scammers and pump-and-dumpers who have since moved on to meme stocks, AI content farming, and so on.

But I wouldn't use crypto as a benchmark because AI has more substance. We can debate if it's going to change the world, but you can build some new types of businesses and services if you have near-perfect natural language comprehension on the cheap.

I know a few people who got wealthy by being early to crypto. None of them had the correct reasoning at the time: They thought BTC was going to become a common way to pay for things or that “the flippening” was going to see worldwide currency replaced with BTC. They thought they’d be kings in a new economy but instead they’re just moderately wealthy with a large tax bill they’re determined to dodge.

I know far more people who lost money on crypto, though. Some were even briefly crypto-rich but failed to sell before the crash or did things like double down on the altcoin bubble.

The second group had gone quiet about their crypto while the few people in the first group gloat and evangelize (because continued evangelization is necessary to keep their portfolios pumped). This creates an intense survivorship bias where it appears like all the crypto kiddies are wealthy, but a quiet mass of people who played with crypto are most definitely not.

I think the framing just doesn't help at all.

I don't understand how this, at all, makes "the point" for anyone.

Anecdata, but the few people I know who were looking to switch gigs all had multiple offers within a few weeks. One thing they all had in common was taking a very targeted approach with their search and leveraging their networks. Not spamming thousands of resumes into the ether.

There’s really not much stopping changing tires from being automated away. Further standardization of tires or wage increases would probably do the trick.

There’s still plenty of software to be created. You’ll probably have to learn some ML tricks or whatever, but there’s nothing going away, just changing as software has always done.

Bad idea. Automotive repair is barely a moat, because you don't need that much training to work those jobs. There's a lot of people who want to do it. And cars are definitely susceptible to recessions - if fewer people are buying cars, if there's a shift to transit, if your locality builds more pedestrian-friendly infrastructure, if businesses that use work vehicles are forced to close, then your demand drops and everyone already in the field is forced to compete with one another.

For moats, look for things that are complex (not everyone can do it), licensed and always needed.

For an old dog like myself it feels an unjust rug pull.

Programmers (and other white collar jobs) were able to luxuriously coast along the ZIRP era because capital (replenished twice via quantitative easing) was cheap and plentiful, and because the elites at the top had to pump huge amounts money to create a shared fantasy of the "technological future" that validates the neoliberal era. Now that the reality of the actual "physical economy" (the economy of making tangible things) has clawed back at us because of that forbidden three-letter word (war), we all realize that doubling and tripling oil prices were actually dictating our lives rather than some "Skynet AI" crap, and thus our fantasy simulacra of "virtual" play-things have now come to an end. Oh and we all found out that most of SaaS was actually bullshit anyway. In fact, if it could be completely replaced by AI then it was already pretty bullshit in the first place.

So, for smart STEM people uninterested in programming and only looking for a stable career, I think they would be better off by just doing engineering work that's a bit more tangible, like robotics, manufacturing, shipbuilding, construction, etc. (Or anything related to war, but only if you're able to stomach what you're doing.) If you don't like to sit all day for a salary, then niche blue collar work can also be a good option, since general-purpose robotics (Physical AI?) is still too far away because of many, many issues that's just too long to explain here. I still think if you like programming then you should stick to it in the long run - there will be a very cold winter because of the combination of LLMs, AI bubble pop, and general economic depression, but for those who survive this era there will be an opportunity because of the shortage of skilled programmers (since no-one bothered to hire juniors after the pop, no one will grow to become seniors themselves!) Computing will still be with us forever, just not in a way that investors thought that it's going to "engulf the world".

You're right, it's possible. But you might be both overestimating the ease of onboarding and underestimating the variety of tasks and constraints devs are responsible for.

I've seen Claude knock out trivial stuff with a sufficiently good spec. But I've also seen it utterly choke on a bad spec or a hard task. I think these outcomes are pretty broadly established. So is the expectation that the tech will get better. Waiting isn't unwise.

To keep and/or increase my current compensation, I have to be competitive in the software development market.

(Whether I need AI to remain competitive is another matter.)

The 16,000 new babies will be competing in different markets.

Oh, and of those 16,000 babies, many are born in far less fortunate circumstances, they're already far behind their cohort. :/

You are giving too much credit to tech journalists. How many of them truly understand Bitcoin or AI?

> Or even more specifically, maximize your foothold in it while minimizing your downside.

Most of my AI usage comes from doing things I don't enjoy doing like making a series of small tweaks to a function or block of code. Honestly, I just levelled the playing field with vim users and its nothing to write home about

Mistakes are less costly in the beginning and the knowledge gained from them is more valuable.

Over-sharing on social media. Secret / IP leaks with LLMs. That kind of thing.

I agree:

FOMO is an all-in mindset. Author admits to dabbling out of curiosity and realizing the time is not right for him personally. I think that's a strong call.

It may have reduced the time to an implementation, based on my experiences I sincerely doubt the veracity of applying the adjective "working".

Sure, maybe crypto changed some lives, but an entire industry? I think ALL of software dev is going under a transformation and I think we're past the point of "wait it out" IMO.

Or I'm wrong, but right I'm being paid to develop a new skill professionally. Maybe the skill ends up not being useful - ok, back to writing code the old way then.

I know, I know. I'm prompting it wrong. I'm using the wrong model. I need to pull the slot-machine arm just one more time.

I know I'm not as clever as Terrance Tao - so I'll wait until the machines are useful to someone like me.

Except you would've probably sold it at any of 1.5x, 2x, 4x, or 10x points. That's what people keep missing about this whole "early bitcoin". You couldn't tell it will 2x at 1.5x, you couldn't tell it will 4x at 2x, and so on.

Yes, they tend to be incredible gullible to certain things, over-simplistic and over-confident but also very "agile" when it comes to sweep their failures under the rug and move on to keep their own neck in one piece. At this point in time even the median CEO knows AI has been way overhyped and they over invested to a point of absolute financial insanity.

The first line of defense about the pressure to deliver is to mandate their minions to use it as much as possible.

We spent a fortune on this over-rated Michelin star reservation, and now you kids are going to absolutely enjoy it, like it or not goddammit!

There are loads of BS tools out there of course but I don’t use that many tools.

I think the logical thing to do is to invest a minor amount of time/money across a broad spectrum of new promising tech. If you had been aware of and bought $500 of Bitcoin in 2010, you'd be a billionaire today. The early people involved with NFTs also did very well.

The Flash example is specifically the opposite of my point. Flash was a lucrative skill for a period of time, but at a certain point it became very clear that it didn't have a future.

Cloud had a very similar vibe when it was really running advertising to CIO/CTOs hard. Everything had to be jammed into the cloud, even if it made absolutely no sense for it to be run there.

This seems to come pretty frequently from visionless tech execs. They need to justify their existence to their boss, and thus try to show how innovative and/or cost cutting they can be.

This is exactly what's happening. The top 5 or 6 companies in the s&p 500 are running a very sophisticated marketing/pressure campaign to convince every c-suite down stream that they need to force AI on their entire organization or die. It's working great. CEOs don't get fired for following the herd.

But is this something that is best done top to bottom, with a big report, counting tokens? Hell no. This is something that is better found, and tackled at the team level. But execs in many places like easy, visible metrics, whether they are actually helping or not. And that's how you find people playing JIRA games and such. My worse example was a VP has decided that looking at the burndown charts from each team under them, and using their shape as a reasonable metric is a good idea.

It's all natural signs of a total lack of trust, and thinking you can solve all of this from the top.

Sometimes it is better to get into things early because it will grow more complex as time goes on, so it will be easier to pick up early in its development. Consider the Web. In the early days, it was just HTML. That was easy to learn. From there on, it was simply a matter of picking up new skills as the environment changed. I'm not sure how I would deal with picking up web development if I started today.

> There are a 16,000 new lives being born every hour. They're all starting with a fairly blank slate.

Not long ago we were ridiculing genZ for not knowing why save icon looks like a floppy disk.

Do you want to feel like that in next 5-10 years?

As a freelancer I do a bit of everything, and I’ve seen places where LLM breezes through and gets me what I want quickly, and times where using an LLM was a complete waste of time.

That's the point of the blog post. If you can't even say right now whether it's for the better, then there's no reason to rush in.

*I should qualify that "using" CC in the strict sense has no learning curve, but really getting the most out of it may take some time as you see its limitations. But it's not learning tech in the traditional sense.

I can't really agree. I've never seen anything from an LLM that I would consider even helpful, never mind transformative.

How are you supposed to use them?

It's clearly a textbook example of survivorship bias.

In the 90s the same argument was directed at this new thing called the internet, and those who placed a bet on it being a fad ended up being forgotten by history.

It's rather obvious that this AI thing is a transformative event in world history, perhaps more critical than the advent of the internet. Take a look at traffic to established sites such as Stack Overflow to get a glimpse of the radical impact. Even in social media we started to see the dead internet theory put to practice in real time.

And coding is the lowest of low hanging fruits.

So just read up on it and say you do. They don't really need 3 years experience, so you don't really need to have it.

lol no. There's nothing actually different about managing VMs in EC2 versus managing physical servers in a datacenter. It's all the same skills, and anyone who is competent in one can pick up the other with zero adjustment.

> But the curious early adopters were the ones best positioned to be leading the charge on "cloud migration" when the business finally pulled the trigger.

From a technological perspective, these sysadmins were right: in nearly all cases (exception: you have a low average load, but it is essential that the servers can handle huge spikes in the load), buying cloud services is much more expensive overall than using your own servers.

The reason cloud computing took of is that many managers believed much more in the marketing claims of the cloud providers than in the technological expertise of their sysadmins.

Obviously there are tons of tools and systems building up around LLMs, and I don't intend to minimize that, but at the end of the day, an LLM is more analogous to a tool such as an IDE than a programming language. And I've never seen a job posting that dictated one must have X number of years in Y IDE; if they exist, they're rare, and it's hardly a massive hill to climb.

Sure, there's a continuum with regards to the difficulty of picking up a tool, e.g. learning a new editor is probably easier than learning, say, git. But learning git still has nothing on learning a whole tech stack.

I was very against LLM-assisted programming, but over time my position has softened, and Claude Code has become a regular part of my workflow. I've begun expanding out into the ancilary tools that interact with LLMs, and it's...not at all difficult to pick up. It's nothing like, say, learning iOS development. It's more like learning how to configure Neovim.

In fact, isn't this precisely one of the primary value propositions of LLMs -- that non-technical people can pick up these tools with ease and start doing technical work that they don't understand? If non-technical folks can pick up Claude Code, why would it be even _kind_ of difficult for a developer to?

So, I'm with the post author here: what is there to get left behind _from_?

Feels like a false dichotomy.

Have I become faster with LLMs? Yes, maybe. Is it 10x or 1000x or 10,000x? Definitely not. I think actually in the past I would have leaned more on senior developers, books, stack overflow etc. but now I can be much more independent and proactive.

LLM-based tools are a wide spectrum, and to argue that the whole spectrum is worth exploring because one sliver of it has definite utility is a bit wonky. Kind of like saying $SHITCOIN is worth investing in because $BITCOIN mooned as a speculative asset:

  - I’m bullish on LLMs chat interfaces replacing StackOverflow and O’Reilly
  - I could not be more bearish on Agents automating software engineering

I still think it's stupid, but I'd be a whole lot richer if I went along with it at the time!

I was ahead of the game with my intimidate expertise in ActionaScript and Silverlight! I made 3D engines in browsers well before WebGL was a spec.

It was quite profitable for a few years, then poof. Dead end lol

> - If you'd started making mobile games when the iPhone was released, you could have built the first Candy Crush

I disagree:

Concerning the first point: how neural networks are today is very different from how they were in former days. So, the knowledge of neural networks from the past does only very partially transfer to modern neural networks, and clearly does not make you a very sought-after specialist right now.

Concerning your second point: the success of mobile games is very marketing-centric. While it is plausible that being early in mobile games development when the iPhone was released might have opened doors in the game industry for you, I seriously doubt whether having this skill would have made you rich.

My goal in life is not to maximize financial return, it's to maximize my impact on things I care about. I try to stay comfortable enough financially to have the luxury to make the decisions that allow me to keep doing things I care about when the opportunities come along.

Deciding whether something new is the right path for me usually takes a little time to assess where it's headed and what the impacts may be.

If you broaden the comparison (only a little bit) it looks suspiciously like employees being forced to train their own replacement (be that other employees, or factory automation), a regular occurrence.

But I think it's just a matter of when not if.

My current guess at my slow fortune 500 is ~1-2 years before we see real employment impact.

Startups are happening now at least with my anecdotal conversations. Right now the discussion is more just slower growth than actually doing layoffs. That coin will flip at some point.

Just finished a search - agree. The resume process is fundamentally broken, but a strong network makes it irrelevant. Lean on connections - there's a ton of opportunity out there.

Or have bet on Mercurial. Which is also close to dead. Or darcs, which has been big in certain environments and now practically extinct.

https://en.wikipedia.org/wiki/Jevons_paradox

You need to study for 3 years to be a car mechanic. And even then you'll need baby sitting for a while in the auto-shop because no one trusts the new guy fresh out of school, with good reason.

>There's a lot of people who want to do it.

No there isn't. What do you mean by lot of people? Automotive repair is a blue collar job with intense physical strain, you're exposed to chemicals you shouldn't be, there will be hearing loss involved no matter how much protection you have and it doesn't even pay all that well, considering the risks involved and the amount of training you need. And no, it's not because the market is saturated with car mechanics, it's because auto repair shops have a lot of pressure to be cheap. Job listings are full of car mechanic openings, you'll never be unemployed.

All the "if's" presented are solved by relocation and not even that much of it, except for this:

>if fewer people are buying cars

Then the skills are easily transferable to other vehicles. But less people are buying cars already, they use uber, which involves a car. A car that needs 10x more repair time than the car you drive daily.

So yeah, auto repair is a good moat. It's complex, not everyone can do it, it's not licensed in most cases (unless you work for a brand or a niche) but there's reputation involved and it's always needed. It just doesn't pay all that well, specially not compared to what I see on HN's monthly whoshiring.

This question is easy to answer in hindsight but it's not trivial to answer in the moment. I like your mindset though

Most execs didn't get where they were by being truly helpful and adding value to the company. They played the game long enough to know that politics trumps accomplishments. The rest from there is the ability to weave a good story (be it slightly or completely exaggerated).

It's not even about trust. It's about incentives in a structure that is dog-eat-dog. Rugged individualism in a corporate structure is a self defeating prophecy. But it's inevitable when executives extract from the company instead of rising the tides for all ships. And shareholders reward it.

I think it depends on why you do programming. I like programming for its own sake. I enjoy understanding a complex system, figuring out how to make change to it, how to express that change within the language and existing code structure, how to effectively test it, etc. I actively like doing these things. It's fun and that keeps me motivated.

With AI I just type in an English sentence, wait a few minutes, and it does the thing, and then I stare out the window and think about all the things I could be doing with my life that I enjoy more than what just happened. I find my productivity is way down this year since the AI push at work, because I'm just not motivated to work. This isn't the job I signed up for. It's boring now.

The money's nice, I guess. But the joy is gone. Maybe I should go find more joy in another career, even if it pays less.

Hopefully not too many people are "enhanced" to the tune of 0.5x!

> But that extra value doesn't accrue to individuals, it accrues to business owners.

What is value?

Is a 2X faster lumberjack 2X as valuable? Sure

Is a 2X faster programmer 2X as valuable? At what, fixing bugs? Adding features? That's not how the "ownership class" would define value.

Productivity is a measure of efficiency, not growth. Slashing labor costs while maintaining the status quo is still a big productivity gain.

guess it's death and destitute for introverts

edit: please explain the downvotes, i'm curious why you think i'm wrong

if what op says it's true, that today only networking works, then it easily follows that if for some reasons you do not have a network then you don't get hired

Sounds like you've never changed a tire. Or at least not outside of a very controlled environment.

These are predictable jobs with very few variables that there is still no sign of automation replacing any time soon. They often don't suck as bad as people think. One of the most enjoyable jobs I had was on an assembly line, because my mind was mostly free to wander. It was almost like meditation.

Our software industry has specialized, for decades, in "rug pulling" / changing / "disrupting" other industries on a massive scale.

I find it pretty ironic when engineers make these statements in that context.

I know many jobs are about giving partial access to secrets or insider knowledge etc but I simply can't see myself accepting that this is my value proposition.

No, let the pie grow. Let more people be able to do more things. Use the new capabilities to do even more. See how you can provide genuine value in the new environment. I know it isn't easy. There are many unknowns. But at least aspirationally I see that as the only positive way forward.

The same thing has happened to many jobs. 100 years ago being a photographer was a difficult skill. They must have felt a rug pull when compact cameras became mainstream and they were no longer called to take all family pictures. Surely the codex writers felt a rug pull when printing became widespread. Typesetters when people could use word processors on their PC with font settings. Prop designers and practical effects people when movies switched to vfx. Etc etc.

It has often been the case for technologies though, like “now we’re doing everything in $language and $technology”. If you see LLM coding as a technology in that vein, it’s not a completely new phenomenon, although it does affect developers differently.

If only Mt Gox didn't vaporize all of my bitcoin in the early 2010's :(

100% accurate - some of us are old enough to have lived through a few of the mini-revolutions in between the mega-revolutions of Internet/Web in the 1990s and now AI/LLM in the 2020s.

We are in the "stupid phase" of adoption still. C-level people have to follow the herd and they are being evaluated on keeping up with everyone else. Idiotic mandates are a way to cause things to happen short-term even though everyone knows long-term it will have to be re-done.

Consultants gonna make a looooooooot of money this coming decade.

S&P 500 Concentration Approaching 50% - https://news.ycombinator.com/item?id=47384002 - March 2026

> No of course there isn't enough capital for all of this. Having said that, there is enough capital to do this for a at least a little while longer. -- Gil Luria (Managing Director and Analyst at D.A. Davidson)

OpenAI Needs a Trillion Dollars in the Next Four Years - https://news.ycombinator.com/item?id=45394071 - September 2025 (8 comments)

I’ve seen people use notepad and I’ve seen people who are so good at vim that they look like they’re on editing code directly with their mind.

Your particular example is extreme and my guess is the coworker is just not great at debugging. I use Claude all the time for finding bugs, but it fails fairly frequently though. I think there’s probably advantage to having some people who don’t use it that often, so you have someone to turn to when it fails.

I’m definitely not exercising my debugging skills as much as I used to and I’m fairly confident they’ve atrophied.

Are the bootcampers better developers? Probably not. But they still were employable and paid relatively the same.

Remember all the hoopla over how people needed be a "prompt engineer" a couple years back? A lot of that alchemy is basically totally obsolete.

Think about the hoops you had to jump through with early GenAI diffusion models: tons of positive prompt suffixes (“4K, OCTANE RENDER, HYPERREALISTIC TURBO HD FINAL CHALLENGERS SPECIAL EDITION”) bordering on magical incantations, samplers (Euler vs. DPM), latent upscalers, CFG scales, denoising strengths for img2img, masking workflows, etc.

And now? The vast majority of people can mostly just describe desired image in natural language, and any decent SOTA model can handle the vast majority of use cases (gpt-image-1.5, Seedream 4, Nano-banana).

Even when you’re running things locally, it’s still significantly easier than it used to be a few years ago, with options like Flux and Qwen which can handle natural language along with a nice intuitive frontend such as InvokeAI instead of the heavily node-based ComfyUI. (which I still love but understand it's not for everybody).

That said, your point about the leverage of learning html and web in the early days compared to now rings true. pre-compiled isomorphic typescript apps are completely unrecognizable from the early days of index.html

Prompt engineering as a specific skill got blown out of proportion on LinkedIn and podcasts. The core idea that you need to write decent prompts if you want decent output is true, but the idea that it was an expert-level skill that only some people could master was always a lie. Most of it is common sense about having to put your content into the prompt and not expecting the LLM to read your mind.

Harnesses isn’t really a skill you learn. It’s how you get th LLM to interact with something. It’s also not as hard as the LinkedIn posts imply.

Mixture of Experts isn’t a skill you learn at all. It’s a model architecture, not something you do. At most it’s worth understanding if you’re picking models to run on your own hardware but for everything else you don’t even need to think about this phrase.

I think all of this influencer and podcast hype is giving the wrong impression about how hard and complicated LLMs are. The people doing the best with them aren’t studying all of these “skills”, they’re just using the tools and learning what they’re capable of.

Next thing I'm waiting on is building a new server for a powerful locally hosted LLM in 5 years. No need to go through the headaches and cost of doing it now with models that may not be powerful enough.

I use LLMs pretty regularly, so I'm familiar with the kinds of tasks they work well on and where they fall flat. I'm sure I could get at least some utility from Claude Code if I had an unlimited budget, but the voracious appetite for tokens even on a trivially small project -- combined with a worse answer than a curated-context chatbot prompt -- makes its value proposition very dubious. For now, at least.

* I considered trying Opus, but the fundamental issue of it eating through tokens meant, for me, that even if it worked much better, the cost would dramatically outweigh the benefit.

If you have try teaching someone something from the absolute ground up, you will quickly realize that a huge number of things you now believe are "standard assumptions" or "obvious" or "intuitive" are actually the result of a lot of learning you forgot you did.

Almost all people are "forgotten" by history.

In any case, people who were not even born yet in the 1990s are using the internet today, very successfully, so clearly you can wait.

Of course don't fraud by like pretending you're a statistician when you have absolutely no mathematical background, but also don't take at face value the "Must have {x} years of experience in {y} tech" requirement when you know you have the necessary work experience to have a good grasp on it in a few weekend prototypes, and you also know that the job doesn't actually require deep expertise of that particular tech.

I did the same for my first React.js job, and I didn't feel bad because 1) I was honest about it and did not sold myself as a React expert, and 2) I had 10 years of front-end development, and I understood web dev enough to not be baffled by hooks and the difference between shallow copy vs. deep copy of a data structure, so passing technical test was good enough for it.

Bikers in the Tour de France used to not wear helmets. They were seen as uncouth (“why jump on the bandwagon?”). Helmets today are way better than they were then. But if the utility provided is greater than the cost, of course it makes sense to act sooner.

I’m not explicitly arguing for investing in AI or other newfangled tech, I’m arguing that the premise of waiting may be “sounded” but also “leaves money on the table”, or in some cases, lives.

The author talks about vaccines as a counter example but doesn’t really address the cost/benefit in any detail.

https://nationalpost.com/news/canada/canada-falls-in-the-wor...

https://financialpost.com/news/canada-manufacturing-longest-...

https://thehub.ca/2026/02/02/canadas-spiralling-debt-problem...

Author’s point is that competitiveness can come in many forms. Having the same AI proficiency as everyone else isn’t differentiating. (And it isn’t table stakes.)

Funny enough, I got laid off last month, yes I’m a tech guy, now they apparently regret it because they are now scrambling to find a replacement to do the tech tasks!

TBH, I’m happy I got laid off because I’m finally building something I wanted to use.

Many developers who picked up the web in the early years struggle with (front-end) web development today. It doesn't matter if they fetched jQuery or MooTools from some CDN as it was done in the mid 00s. Once the tooling became too complicated and ever changing they couldn't keep up as front-end dilettante. It required to commit as professionals.

If you started today, you'd simply learn the hard way, as it's always been done: get a few books or register for a course. Carve some time every day for theory and practice. All the while prioritizing what matters the most to get stuff done quickly right now, with little fluff. You will not learn Grunt, Bower, and a large array of historic tech. You'll go straight for what's relevant today. That applies to abstractions, frameworks, and tooling, but also to the fundamentals. You'll probably learn ES6+ and TS, not JS WAT. A lot of the early stuff seems like an utter waste of time in retrospect.

This is true for all tech. If you knew nothing about LLMs by the end of this year, you could find a course that teaches you all the latest relevant tricks in 5 to 10 hours for 10 bucks.

If you started early webdev, you learned lots of tricks, that dont benefit a modern webdev. E.g soap, long polling, the JsonP workaround... and so on

Many of the Llm frameworks will be seen simular. Mcp is already kinda heading in the obsolete direction imo, as skills took over

My last job was a cable technician - making house calls to fix wifi, satellite tv, phone issues. Mostly elderly residents. The majority of them all were computer and phone illiterate. They were slow adopters to the fast-moving technology and many of them did not know how to operate their devices after we (UI/UX/hardware/software engineer 'we') removed them.

I wonder if this also has contributed to the elderly lonliness problem - sure its probably mostly related to physical companionship, acceptance of aging, etc, but the world that they knew (in general and the technological world they grew up in) is no longer recognizable.

Building a simple marketing website? Probably don’t waste your time - an LLM will probably be faster.

Designing a new SLAM algorithm? Probably LLMs will spin around in circles helplessly. That being said, that was my experience several years ago… maybe state of the art has changed in the computer vision space.

Without playing around with it, you wouldn't know when to use an LLM and when not.

If you test specific features of those solutions over time you see very inconsistent results, lots of lies, and seemingly stable solutions that one-shot well but suddenly experience behaviour changes due to tweaks on the backend. Tuesdays awesome agent stack that finally works is loading totally different on Thursday, and debugging is “oh, sorry, it’s better now” even when it isn’t. Compression, lies, and external hosting are a bad combo.

Sometimes I imagine a world where computers executed programs the same way each time. You could write some code once and run it a whole calendar month later with a predictable outcome. What a dream, we can hope I guess.

Kind of weird tools also incorporate addictive gambling game's UX design. They're literally allowing you to multiply your output: 3x, 4x, 5x (run it 5 times for a better shot at a working prompt). You're being played by billionaires who are selling you a slot machine as a thinking machine.

There isn't? Then why is it that whenever devs have tried it and not achieved useful results, they're told that they just haven't learned how to use it right?

Projects as simple as "set up a tmux/vim binding so I can write prompts in one pane and run claude in the other". Fails.

I've been coding for over 20 years.

If there is no learning curve, why doesn't it work for me? You can't say I'm not using it right, because if that was true, then all I need to do is climb the learning curve to fix that, the curve that you say doesn't exist.

I am conservative regarding AI driven coding but I still see tremendous value.

It makes me want to ask you: do you ever see helpful things from your colleagues at all?

"It's rather obvious that this AI thing is a transformative event in world history" perhaps but it's not at all obvious how it's going to shake out or which bets are sensible.

It's worth noting that SO was declining well before ChatGPT launched. It seems more likely that the decline of SO was more driven by Google ranking changes to prioritise websites that served Google ads. Certainly I remember having to go down a few results to get SO results for a while, even when the top results were just copypasta from SO.

Allow me to introduce you to the dot-com boom, where everyone who bet on the internet went broke.

But something tells me you won't do that.

Who writes software and doesn't have a list of "I'll fix this one day" issues as long as their arm?

This is honestly one of the things I enjoy most at the moment. There's whole classes of issues where I know the fix is probably pretty simple but I wouldn't have had time to sort it previously. Now I can just point claude at it and have a PR 5mins later. It's really nice when you can tell users "just deployed a fix for your thing" rather than "I've made a ticket for your request" your issue is on the never-ending backlog pile and might get fixed in 5 years time if you're lucky.

This started literally two weeks ago and a couple of days ago I talked to one of the admin people who wanted an update on the progress I'd made with sanding off some of the rough edges of the very rough implementation that the managing partner had put in place (he bought a Mac Mini, put OpenClaw on it, then gave it admin access to a whole pile of stuff!) I said I needed a couple more days. "Okay," she said, "but I need this quickly, because we're firing people next week."

They have literally gone from no agentic AI, to discovering OpenClaw, to firing people, in a two-week time span.

When economists say that the predicted job losses as a result of AI have not yet shown up in the data, I'm genuinely befuddled. Either we don't have long to wait to start seeing them, or there's something wrong with the data, because you can't tell me what I just described above is an isolated phenomenon.

I also have to say: I've always enjoyed working with this client, but this experience has been a huge turnoff on a number of different levels.

I bet we could replace nearly all the CEOs in the country with chatgpt controlling a ceo@thatcompany.com email and nobody would notice.

Unfortunately that doesn't change my outlook on where all this is headed.

Maybe I didn't express myself properly, but I think we agree, at least on this point?

Besides this effect, of enabling smaller teams to produce the same results, I think there is a larger effect coming where fundamentally different structures produce the same or better results as last year. I just don't think we've completely figured out what that looks like yet.

I have even replaced car tires before and yet still have this opinion.

If you think any programming task at hand one must have at least some reasonable grasp of formalism, boolean logic, predicate logic, then understanding the software developing concepts, your APIs frameworks, language constructs etc and finally the domain knowledge.Most of this goes away when changing from coding to prompting.

I was just doing some computer graphics work myself doing Signed Distance Fields and Claude just literally regurgitated code that I could just adopt (since it works) without understanding any of the math involved.

I'd say that prompting is at least two orders of magnitude easier than coding.

Boy I can't wait for the equivalent of low effort high volume clickbait to take over software. Yay!

Digital products such as "photoshop" have had value because people need a tool like that and there's only a limited number of competition, i.e. scarcity. The scarcity exists because of the cost. I.e. the cost of creating "photoshop"creates limit for how many "photoshops" exist. When you bring down the cost you'll have more "photoshops" when you have more "photoshops" as the volume increases the value decreases. Imagine if you can just tell claude "write me photoshop", go take a dump and come back 30 mins later to a running photoshop. You wouldn't now pay 200USD for a license, now would you? You'd pay 0USD.

If you now create a tool that can (or promises it) can obliterate the costs, it means essentially anyone can produce "photoshop". And when anyone can do it it will be done over and over and at which point they're worth zero and you can't give them away.

The same thing has happened to media publishing, print media -> web, computer games etc.

Then the problem is that when your product is worth zero you can no longer make a business by creating your product, so in order to survive you must look into alternative revenue streams such as ads, data mining etc. None of which are a benefit to to the product itself.

Interestingly, the model doesn't "know" that it's ignoring you. From its perspective, it has retrieved a "meaningful" pattern—virtual parameter names that probably fit common conventions it saw during training. Your actual request simply... wasn't documented.

Not quite on topic but as an engineering manager responsible for IDE development, explaining to recruiters and candidates I wanted engineers who developed IDEs, not just used them. Unfortunately, that message couldn't get through so I saw many resumes claiming, say 5 years of Eclpse experience, but I would later determine they knew nothing of the internals of an IDE.

Presumably, people now claim 3 years of machine learning experience but via ChatGPT prompting.

The question isn’t if you’ve improved. It’s if the path you took to getting to your current improvement could have been shortcut with the benefit of hindsight. Given the number of dead ends we’ve traversed, the answer almost certainly is yes.

I think these ideas are similar to long-term relationships. Identify when it's clear it's worth your time, like the author, and commit appropriately, and then when it's time to move on move on.

AngularJS, Backbone, Knockout, YUI, were all a wave of pretty groundbreaking frontend technology. It was absolutely worth experimenting with and committing to once they had some uptake, but probably not before then unless you wanted to work on the teams building them. Time went on, they had years of longevity that overlapped with the next wave of Vue, React, and the rest, and those became worth investing in long-term. Along the way, fundamentals in underlying web technologies were crucial, programming, logic, networking, markup, design.

Actionscript was totally worth investing in, until it ran it's course, and then other things came along and you would have adapted your game programming and engine programming skills yo a different platform.

If you had worked on anything other than a transformer based architecture post 2016, such as Mamba or RWKV, you would have wasted your time.

Mamba 3 is the third iteration and somehow I doubt that it will catch on.

But there’s some stuff that I don’t bother explore in depth because my time is finite and I don’t really need it. And anything LLM tooling is probably easier than a random JS framework. Vim’s documentation is probably longer than cursor’s.

This is actually where I would be most reluctant to use an LLM. Your website represents your product, and you probably don’t want to give it the scent of homogenized AI slop. People can tell.

OTOH, tfa specifically said:

> I feel the same way about the current crop of AI tools. I've tried a bunch of them. Some are good. Most are a bit shit. Few are useful to me as they are now. I'm utterly content to wait until their hype has been realised.

So, it's not like he's being deliberate ignorant, rather simply deliberately slow-walking his journey.

That's what's being asked of me in my last two jobs. Vibe code it, if it's bad just throw it away and regenerate it because it's "cheap". The only thing that matters is that you can quickly generate visible changes and ship it to market.

Out of frustration I asked upper management (in my current job), if you want me to use AI like that then I'll do it. But when it inevitably fails, who is responsible? If there's no risk to me, I will AI generate everything starting today, but if I have to take on the risk I won't be able to do this.

Their response was that AI generates the code, I'm responsible for reviewing it and making sure it's risk free. I can see that they're already looking for contractors (with no skin in the game) that are more than willing to run the AI agents and ship vibe code, so I'm at a loss on what to do.

I think a decent place to start is: given a small web app, give it a bug report and ask it what causes the bug.

If people put down the AI, and actually learn how to write a `for` loop, they would be more hire-able than 50% of candidates.

> "Guess it's death [...] for introverts"

There is a meritocracy somewhere in our capitalist system. Not everyone participates, but it exists.

This is really just a reversion to how things used to work, relying on human connections. People seemed to manage to get jobs 30 years ago just fine

How are these put on in the first place on an assembly line?

Do you think it's fair that when the society moves underneath, the capitalistic system moves its tectonic plates it's the individual who has to bear the cost of that?

Abd let's be clear only software devs are just sucking it up. You think lawyers and doctors would allow themselves to be laid off en masse and be replaced with trainees who just prompt the computer?

Also what will happen when high wage earners start loosing their discretionary income. The whole service sector for starters will be shaken.

Just imagine some big tech company laying off 10k engineers. Making 0.3m per year. That's 3b dollars that disappear from the incomes and thus from the economy and just stays in the pockets of the capital holders.

Theres a reason most people want a white collar job and send their kids to college instead of to such manual jobs.

That's an incredibly uncharitable reading of the parent comment. At no point in history prior to maybe this year could you argue that working in software was gatekeeping, toll extracting, or rent seeking. Being a highly skilled craftsperson creating software for those who can't or don't want to is a very psychologically positive self identification. Lamenting that the industry is moving away from highly skilled craftspeople is also perfectly valid, even if you believe that it is somehow good for society, which is yet to become clear.

So? Demand the source code. Run your own AI to review the quality of the code base. The contracting company doesn't want to do it? Fine, find one that will.

Now you can make a perfectly tailored resume, apply to 50 jobs in a day, and it's not unexpected to not get any response from those in 2 weeks. You don't know if it's your resume, the company, or the economy. And no one wants to admit the latter two are problems.

Not to mention the utter disrespect these days. There's no decorum in many of these "professional" settings, when normally you want your interview process to show off your best face.

I truly believe so much of the anti-AI sentiment is the same as the Luddites.

They're often used as a meme now, but they were very real people, faced with a real and present risk to their livelihoods. They acted out of fear, but not just irrational fear.

AI is the same: it's unquestionably (to anyone evaluating it fairly) a huge boost to productivity ... and also, unquestionably, a threat to programmer jobs.

Maybe the OP is right about waiting, but to me whenever new tech is disrupting jobs, that seems like the best time to learn it. If you don't, it's not just FOMO as the author suggests ... it's failing to keep up on the skills that keep you employed.

What you aim for if you want to invest early is rather a probability distribution of

- get rich with a small (but nevertheless realistic) probability p

- get something between little, nothing, and loosing a little bit with probability 1-p

This is a very different offering than the profit probability distribution that index funds give you.

Otherwise you would most likely have sold during one of huge crashes or values, attempted trade and lost it all, invested into the new shitcoin NFT or whatever or just got hacked along the way.

Or you could take what's in the box!!

I bought 10ish BTC at some point for almost nothing, sold them for a low 4-digit amount thinking they were stupid anyway. I still think they were stupid but it turns out they could have paid off my house easily. Oh well.

Didn't pull the trigger. I just tell myself I'd have sold them when they doubled in price or they'd have been hacked in one of the mt. gox attacks and I'd have lost them anyways.

Today it would be about 120m. Oh well.

In the vast majority of cases, financial returns help maximize your impact on the things you care about. Arguably in most cases it's more effective for you to provide the financing and direction but not be directly involved. That's why the EA guys are off beng quants.

The only real exceptions are things that specifically require you personally, like investing time with your family, or developing yourself in some way.

The best professionals did not fall for insanity of the modern front-end dilettante and continued hacking shit without that insanitity.

> You will not learn Grunt, Bower, and a large array of historic tech. You'll go straight for what's relevant today.

which will be outdated "tomorrow" just like grunt/bower... are looked at today

> A lot of the early stuff seems like an utter waste of time in retrospect.

This cannot be further from the truth, if you learned Javascript early, like really learned it, that mastery gets you far today. The best front-end devs I know are basically Javascript developers, everything else is "tech du jour" that comes and goes and the less of it you invest in the better off you'll be in the long-run.

> If you knew nothing about LLMs by the end of this year, you could find a course that teaches you all the latest relevant tricks in 5 to 10 hours for 10 bucks.

Hard disagree with this unless you are doing simple CRUD-like stuff

Being a good professional developer means getting the primitives and the data model not horribly pointed in the wrong direction. So it's extremely helpful to be aware of those primitives. And the argument "nobody is better off knowing assembly as a primitive" doesn't hold because as-said the web is literally still html files. It's right there in the source.

My mother has a phone, but only use it to call. She has never needed a computer even though I spent my teenage years glued to one. But I have like 1 percent of a skill in cooking.

I've been impressed by how this isn't quite true. A lot of my coding life is spent in the popular languages, which the LLMs obviously excel at.

But a random dates-to-the-80s robotics language (Karel)? I unfortunately have to use it sometimes, and Claude ingested a 100s of pages long PDF manual for the language and now it's better at it than I am. It doesn't even have a compiler to test against, and still it rarely makes mistakes.

I think the trick with a lot of these LLMs is just figuring out the best techniques for using them. Fortunately a lot of people are working all the time to figure this out.

Yes, it's hard to see how, at this moment in time, "Anybody can write code with an LLM" is so different from "Anybody can make money in the stock market."

The underlying mechanisms are completely different, of course, and the putative goal of the LLM purveyors is to make it where anybody really can write code with an LLM.

I'm typically a nay-sayer and a perfectionist, but many not-so-great things become and stay popular, and this may fall into that category.

> Kind of weird tools also incorporate addictive gambling game's UX design.

It's unclear it started out this way, but since it's obviously going this way, it is certainly prudent to ask if some of this is by design. It would presumably be more worrisome if there were only a single vendor, but even with multiple vendors, it might be lucrative for them to design things so that "true insider knowledge" of how to make good prompts is a sought-after skill.

I'm not sure why it isn't working for you. Maybe your expectation is a perfect one-shot or else it has zero value, and nothing in between?

But my advice is to switch gears and see the "plan file" as the deliverable that you're polishing over implementation. It's planning and research and specification that tends to be the hard part, not yoloing solutions live to see if they'll work -- we do the latter all the time to avoid 10min of planning.

So, try brainstorming the issue with Claude Code, talk it through so it's on the same page as you, ensure it's done research (web search, docs) to weigh the best solutions, and then enter plan mode so it generates a markdown plan file.

From there you can read/review,tweak the plan file. Or have it implement it. Or you implement it. But the idea is that an LLM is useful at this intermediate planning stage without tacking on additional responsibilities.

I think by "no learning curve" they are referring to how you can get value from it without doing the research you'd need to use a conventional tool. But there is a learning curve to getting better results.

I learned my plan file workflow just from Claude Code having "Plan Mode" that spits out a plan file, and it was obvious to me from there, but there are people who don't know it exists nor what the value of it is, yet it's the centerpiece of my workflow. I also think it's the right way to use AI: the plan/prompt is the thing you're building and polishing, not skipping past it to an underspecified implementation. Because once you're done with the plan, then the impl is trivial and repeatable from that plan, even if you wanted to do the impl yourself.

I'm way past the point of arguing anything here, just trying to help.

Because LLMs are not actually good at programming, despite the hype.

No, not at all. I may be using it wrong.

I put in "write me a library that decodes network packets in <format I'm working with>" and it had no idea where to start.

What part of it is it supposed to do? I don't want to do any more typing than I have to.

I think you are missing the point, and also the very site you're posting on.

Look at the top 50 list of most valuable companies in the world. Over half of the total market value reported today is attributed to companies which were either dotcom startups or whose growth was driven by the dotcom growth period. Dismissing the advent of the internet as anything short of revolutionary is disingenuous, no matter how many zombo.com companies failed.

LLMs have the exact same transformative impact on humanity.

They had to hire a bunch of them back less than two months later. The speed-ups were approximately nil and making the editors edit AI slop all day long had them all close to quitting.

They didn't even wait to see if there were any actual benefits, they just blindly fired a bunch of people based on marketing lies. I can only assume they're the same sorts who fall for Nigerian Prince scams.

I think it's important to know and practice your passion, even if you have to work on something different to pay the bills. You can only be good at something if you really like it, and you never know what opportunity you'll stumble onto if you're ready for it.

Patric Boyle has a video on this in case you care for the details.

And ideally a sample large enough to capture any wasted time from dead ends in other tasks where the tool may actually fail to solve the problem.

I’ve definitely lost a couple hours here and there from when it felt like I was right on the verge of CC fixing something but never actually got there and finally had to just do it myself anyway.

> It isn't hard.

you're not an introvert then

I don't think that's it. SO was the go-to page for troubleshooting, whose traffic was not exactly originating from web search. Also, the LLM-correlated drop in traffic is also reported by search engines. Stack Overflow just so happens to be a specialized service with a very specialized audience whose demand is perfectly dominated by LLM chatbots.

I do agree that the notion of difficulty needs to be recalibrated though, seemingly impressive RE tasks can now be done trivially with LLMs.

Most humans are not intrinsically intelligent: they are regurgitators and mimickers of others. Very few think independently and are capable of original thought.

Come at me with the down voting.. I dont care. History will show Im correct.

of course if the process doesn't involve networking then we don't have a problem, we agree on that

If your CEO doesn't look like a taxi dispatcher he's just moving his wings around waiting for a food pellet.

Im working on building something to address this. That's all I'll say lol.

I judge "failing to keep up" by my ability to "catch up". Right now if I search for paying courses on AI-assisted coding, I get a royal bunch for anything between 3$ to about 25$. These are distilled and converging observations by people who have had more time playing around with these toys than me. Most are less than 10 hours (usually 3 to 5). I also find countless free ones on YouTube popping up every week that can catch me up to a decent bouquet of current practices in an hour or two. They all also more or less need to be updated to relevancy after a few months (e.g. I've recently deleted my numerous bookmarks on MCP).

Don't get me wrong, LLM-assisted coding is disruptive, but when practice becomes obsolete after a few months it's not really what's keeping you employed. If after you've spent much time and effort to live near that edge, the gap that truly separates you from me in any meaningful way can be covered in a few hours to catch up, you're not really leaving me behind.

This is all very tiring and difficult. You can be significantly better than other people at this skill.

  AI is the same: it's unquestionably (to anyone evaluating it fairly) a huge boost to productivity .

It’s ridiculous double standard and poisons any reasonable discussion to assert something is a fact and anyone who disagrees is a hysterical Luddite based on no actual evidence.

But even $100 would have been nice given you could still pop them out for free on a standard PC back then with mining software.

Or in prison for fraud.

A lot of snark aside there's a bit of a false dichotomy (I think) here at work. Whenever or wherever your jumping in point is into $something it will always pay dividends to learn the fundamentals of that $something well and unless you interact with older iterations on that $something then you'll never have to bother learning the equivalent of Grunt, Gulp, Stylus, Nunjuncks and so on for that $something.

With that being said it's also good to put aside time once a year to check out a good recommended (and usually paid) course from an established professional aimed at busy professionals.

As for LLMs I feel it's slowly becoming a thing big enough where people will have to consider where to focus their energy starting with 2027. Kinda like some people branched from web development into backend, frontend and UI/UX a good while back. Do you want to get good at using Claude Code or do you want to integrate gen AI features at work for coworkers to use or customers/users? It's still early days just like when NodeJS started gaining a lot of traction and people were making fun of leftpad.

"Front-end professional" and "no tooling" have been exclusive propositions since the early 2010s. You either learned to use tools or you were out of the loop.

> which will be outdated "tomorrow" just like grunt/bower... are looked at today

Not really. Historically, the main problem with front-end development has not been change, but the pace of it. That's how it ties in with the current discussion regarding the (now) ever-changing terrain of LLM-assisted coding. Front-end development is still changing today, but it's coalescing and congealing more than it's revolving. The chasms between transitions are narrowing. If you observe how long Webpack lasted and familiarity with it carried over to using Vite, it's somewhat safe to expect that the latter will last even longer and that its replacement will be a near copy. Someone putting time to learn front-end skills today might reap the benefits of that investment longer.

> if you learned Javascript early, like really learned it, that mastery gets you far today.

I did. I got a copy of the Rhino book 4th ed. and read it cover to cover. I would not advise to learn JS today with historical references. JS was not designed like most other languages. It was hastily put together to get things done and it had a lot of "interesting", but ultimately undesirable, artifacts. It only slowly turned into a more sensible standard after-the-fact. Yes, there are some parts that are still in its core identity, but a lot in the implementation has changed. Efforts like "Javascript: The Good Parts", further standardization, and TS helped to slowly turn it into what we know today. You don't need to travel back in time for that mastery. Get a modern copy of the Rhino book and you'll be as good as the best of them.

I was looking at trying to remember/figure out some obscure hardware communication protocol to figure out enumeration of a hardware bus on some servers. Feeding codex a few RFC URLs and other such information, plus telling it to search the internet resulted in extremely rapid progress vs. having to wade through 500 pages of technical jargon and specification documents.

I'm sure if I was extending the spec to a 3.0 version in hardware or something it would not be useful, but for someone who just needs to understand the basics to get some quick tooling stood up it was close to magic.

The question relevant for LLMs would be "how many high quality results would I get if I googled something related to this", and for DICOM the answer is "many". As long the that is the case LLMs will not have trouble answering questions about it either.

Why would I do that? Well, I wanted to understand more deeply how differences in my prompting might impact the outcomes of the model. I also wanted to get generally better at writing prompts. And of course, improving at controlling context and seeing how models can go off the rails. Just by being better at understanding these patterns, I feel more confident in general at when and how to use LLMs in my daily work.

I think, in general, understanding not only that earlier models are weaker, but also _how_ they are weaker, is useful in its own right. It gives you an extra tool to use.

I will say, the biggest findings for "weaknesses" I've found are in training data. If you're keeping your libraries up-to-date, and you're using newer methods or functionality from those libraries, AI will constantly fail to identify with those new things. For example, Zod v4 came out recently and the older models absolutely fail to understand that it uses some different syntax and methods under the hood. Jest now supports `using` syntax for its spyOn method, and models just can't figure it out. Even with system prompts and telling them directly, the existing training data is just too overpowering.

This is an industry that requires continuous learning.

This is exactly the workflow that works very well for me in Cursor (although I don't use their Plan Mode - I do my version of it). If you know the codebase well this can increase your speed/productivity quite a bit. Not trying to convince naysayers of this, their minds are already made up. Just wanted to chime in that this workflow does actually work very well (been using it for over 6 months).

I've been reading a book about the history of math and at some points in the beginning the author pointed out how some fields undergo a radical change within due to some discovery (e.g. quantum theory in physics) and the practitioners in that field inevitably go through this transformation where the generations before and after can't really relate to each other anymore. I'm paraphrasing quite a bit though so I'll just recommend people check out the book if they're interested: The History of Mathematics by Jacqueline Stedall

And the aforementioned VS Code video, if I remember correctly: https://youtu.be/dutyOc_cAEU?si=ulK3MaYN7_CPO76k

But this is begging the question.

Yes, we can see that the internet was radically transformative.

But you are arguing that this somehow proves that LLMs are too, when there's wildly insufficient evidence—either on where LLMs are going in themselves, or in the comparison—to credibly make that claim.

"Oh, you think I've never changed a tire? Well here is my abstract high level understanding of the steps to changing a tire! And have you considered the quintessential controlled environment for putting tires onto cars?"

Part of the reason for my prior comment is the clear fact that a not-insignificant percentage of white collar jobs are being massively devalued at the moment, which means many people who thought they'd be able to send their kids to college with income from such jobs won't.

Considering that the field of robotics is so far behind LLMs in terms of clear value outside of niche industrial applications, I think manual labor is about due for a resurgence. There may be some major rebalancing happening. The big question for laborers will be - as it has always been - what can I do that sucks the least but also allows me to pay for a decent life? Answers will vary.

Yes, producing software was value. (It of course still is as of today, we are talking about what may be coming). My plead is to continue searching for ways to contribute value. Don't resign to a feeling that the only way to hold on is if you try to stop others from knowing about or being able to use the skill leveling tech. This makes one bitter and negative. Embrace it, aspire to be happy about it.

Its like getting scooped in science. In research, I always try to reframe it to be happy that science has progressed. Let me try to learn from it and pivot my research to some area where I can contribute something. Sulking about having been scooped does not lead to positive change and devalues ones own self-image.

To paraphrase another analogy that I enjoyed, it’s a bit like when 3d printing became a thing and hype con artists claimed that no one would buy anything anymore, you could just 3d print it.

Sure. I’m saying someone pursuing that portfolio will probably end up underperforming an index. Most new early-stage VCs do.

> get something between little, nothing, and loosing a little

Broadly speaking, when your investment outcomes don’t differentiate between anything and zero, you’re mostly going to get zero.

There was a local food delivery service at the time that accepted bitcoin. Can you imagine looking back on life and realizing you spent the equivalent of $1M on a burrito?

I've not found this to be true at all, for a variety of reasons. One of my moral principles that extreme wealth accumulation by any individual is ultimately harmful to society, even for those who start with altruistic values. Money is power, and power corrupts.

Also, the further from my immediate circle I focus my impact on, the less certainty I have that my impact is achieving what I want it to. I've worked on global projects, and looking back at them those are the projects I'm least certain moved the needle in the direction I wanted them to. Not because they didn't achieve their goals, but because I'm not sure the goals at the outset actually had the long term impact I wanted them to. In fact, it's often due to precisely what we're talking about in this thread: sometimes new things come along and change everything.

The butterfly effect is just as real with altruism as it is with anything else.

So, the things that matter the most for most people?

Studies pretty consistently show that happiness caps off at relatively modest wealth.

People don't become quants because they are EAs, they become EAs to justify to themselves why they became quants.

The EA guys aren't the final word on ethics or a fulfilling life.

Ursula K. Le Guin wrote that one might, rather than seeking to always better one's life, instead seek to share the burden others are holding.

Making a bunch of money to turn around and spend on mosquito nets might seem to be making the world better, but on the other hand it also normalizes and enshrines the systems of oppression and injustice that created a world where someone can make 300,000$ a year typing "that didn't work, try again" into claude while someone else watches another family member die of malaria because they couldn't afford meds.

93% of Developers Use AI Coding Tools. Productivity Hasn't Moved. - https://philippdubach.com/posts/93-of-developers-use-ai-codi... - March 4th, 2026

A very simple kind of query that in my experiences causes problems to many current LLMs is:

"Write {something obscure} in the Wolfram programming language."

> The web is special in this sense, it's intentionally long-lived warts and all. So the fundamentals pay outsized dividends.

Fundamentals pay dividends, but what makes you think that what you learn as an early adopter are fundamentals? Fundamentals are knowledge that is deemed intemporal, not "just discovered".

The historical web and its simplicity are as available to anyone today as it was back then. People can still learn HTML today and make table-based layouts. HTML is still HTML, whether you learned it then or today. But if back then you intended to become a professional front-end developer, you would still have to contend with the tremendous difficulties that some seem to have forgotten out of nostalgia. You'd soon have to also learn CSS in its early and buggy drafts, then (mostly non-standard) JavaScript (Netscape and IE6) and the multiple browser bugs that required all kinds of hacks and shims. Then you'd have to keep up with the cycles of changing front-end tools and practices, as efforts to put some sense into the madness were moved there. Much in all that knowledge went nowhere since it was not always part of a progression, but rather a set of competing cycles.

Fundamentals are indisputably relevant, but they're knowledge that emerges as victorious after all the fluff of uncertainty has been left behind. Front-end development is only now settling into that phase. With LLMs we're still figuring out where we're going.

Even if your architectural idea is completely unique... a never before seen magnum opus, the building blocks are still legos.

For a single dev team, vibe coding is great. Write specs, write plans, write code. I know what the project wants and needs because I'm the target market.

At work, I haven't written more than a few lines of code since December. But I work with other people vibe coding this same project. Lots of changing requirements and rapid iteration. Lots of mistakes were made by everyone involved. Lots of tech debt. Sure, we built something in 2 mos that would have otherwise taken us 6 mos, but now I'm fixing the mess that we caused.

I think the critical difference is the attitude towards our situation. My boss said to fix the AI harness so we can vibe code more confidently and freely. But other bosses might cut their losses and ban vibe coding. Who's right? I dunno. In both cases I'd just do what my boss wants me to do. But it's not that I don't want to be left behind. I don't want to lose my job. There's a difference.

Why? Because all the folks involved have created a technology in search for a problem to solve. That never, ever works. Steve Jobs of all people left this piece of wisdom behind. Its amazing how few actually apply it.

The internet was never this - its origins go back to the need to able to transmit data - darpa. And this is what we still do now...

If people really counted all the time they spend coddling the AI, trying again, then trying again and again and again to get a useful output, then having to clean up that output, they would see that the supposed efficiency gains are near zero if not negative. The only people it really helps are people who were not good at coding to begin with, and they will be the ones producing the absolute worst slop because they don't know the difference between good and bad code. AI is constantly trying to introduce bugs into my codebase, and I see it happening in real-time with AI code completion. So, no you aren't "holding it wrong", the other people are no different than the crypto-bro's who were pushing blockchain into everything and hoping it would stick.

Where LLMs are behind humans is depth of insight. Doing anything non-trivial requires insight.

The key to effectively using LLMs is to provide the insight yourself, then let the LLM do the grunt work. Kind of like paint by numbers. In your case, I would recommend some combination of defining the API of the library you want yourself manually, thinking through how you would implement it and writing down the broad strokes of the process for the LLM, and collecting reference materials like a format spec, any docs, the code that's creating these packets, and so on.

Is using an LLM the same as writing JavaScript over Assembler? idk

I guess it's the same argument of doing math yourself vs. using a calculator, gets the job done

But yeah it goes back to my perspective of why be a carpenter/furniture maker when a 3D printer can just spit one out

Why milk the cow when you can just buy the milk

On the other hand, I can see these tools getting good enough that scope creep doesn't even matter.

ATM I usually get stuck around the review/verification stage. As in, my code works, I have tested that it works, but it is failing CI or someone left a PR comment. And for each comment I'll have to make sure it makes sense, make the change, test again, and get CI passing again.

Shareholders only care about short term gains, CEOs have no skin in the game, everyone else under wants to keep their job. None of these are aligned towards "make the nest proudct and satisfy customers".

That is not at all what I said. Please do not misrepresent.

I said they took a targeted approach *and* exercised their networks. Those are two separate things.

I'm also daydreaming about other careers instead of doing something useful.

Anyways, checking happens often enough that the risk of being considered a liar and a fraud for claiming experience you don't have is high.

Admirable self-reflection.

Your first paragraph is just a standard response to utilitarianism, although a poor one because it doesn't consider EV.

Nonetheless I'm not quite sure why merely mentioning EA draws out all these irrelevant replies about it. It was incidental, not an endorsement of EA.

And to be fair, Steve Jobs was a master of taking things that had been invented elsewhere, and making them work well enough to foster a demand.

But your point stands. Who made the most money, Xerox PARC, or Apple?

I don't agree. It can't write code at all, it can only copy things it's already seen. But, if that is true, why can't it solve my problem?

> The key to effectively using LLMs is to provide the insight yourself, then let the LLM do the grunt work

Okay, so how do I do that? Remember, I want to do ZERO TYPING. I do not want to type a single character that is not code. I already know what I want the code to do, I just want it typed in.

I just don't think AI can ever solve a problem I have.

Actually writing code is the fun and easy bit.

So then you have no choice but to seek alternative revenue streams (ads, data mining) and in fact this becomes the thing, since the original thing no longer produces a revenue.

But also, a lot of the manual labor is quite expensive and only affordable as long as there are white collar workers who can pay for fancy bathroom remodelings and landscaping and so on. I don't know how a big deluge of reskilled pipefitters and HVAC technicians will be able to find work. Will everyone just pay each other to do a bunch of handy work for each other?

We're about to pull the rug underneath all knowledge workers. This will disrupt wage earners lives. This will disrupt the economy.

You might feel great about when things become cheaper but remember that when things are cheap it's only because costs are low and when costs are low the revenues are low and when revenues are low salaries are low too. Keep in mind that one party's cost is other party's revenue.

The economy is ultimately one large circle where the money needs to go around. You might think of yourself a winner as long as someone else's salary drops to zero and you still get to keep your income but eventually it will be you whose income will also be disrupted.

Just something to keep in mind.

And also we're going to just not rug pull on the individual knowledge workers but businesses too. Any software company with a software product will quickly find themselves in a situation where their software is worth zero.

Also this comment about gatekeeping is absolutely stupid. It's like saying trained doctors and medical schools are gatekeeping people from doctoring. It would be so much better if anyone could just doctor away, maybe with some tool assistance. So much fantastically better and cheaper? Right! Just lay off those expensive doctors and hire doctor-prompters for a fraction of the price.

One person hand coding and one person having Claude still results in the same output that is compatible with each other.

This is more like mandating that you use vim. I’ve never some something like that before in 20+ years.

This holds if you consider "underperforming" to be a comparison of expected values.

On the other hand, if you consider "probability of getting a really huge payoff" to be the measure by which the investments are compared, the index fund is the one that looses in the comparison.

I wasn't able to replicate in my own testing though. Do you know if it also fails for "mathematica" code? There's much more text online about that.

Picking up the web early didn't help with the latter. I spent most of my early time memorizing tips and tricks that only applied to old browsers. I didn't pick up the fundamentals till I went back to school for CS and took a networking class.

You're right, fundamentals are distilled, so to think they are free just by getting in early is likely backwards. And earning one's professional chops doesn't stop or start based on when you enter.

Web dev definitely is nostalgic. I miss the early days but I also conveniently erased ie6, binding data to HTML, the need for backbone and jQuery to do anything. hmmm yeah doesn't matter when you start, it's all a grind if you dig deep enough.

You didn't actually build it in 2 months.

Exactly. I counted and reported my results in a previous thread [0].

[0] https://news.ycombinator.com/item?id=47272913

If you were the type of person who makes tiny toy apps, or you worked on lots of small already been done stuff, you'd love doing this. It would speed you up so much.

But if you worked on a big application with millions of users that had evolved into it's own snowflake through time and use, you'd get very little from it.

I think I probably could benefit from looking at existing open source solutions and modifying them a lot of the time, and I kinda started out doing that at first. But eventually you realize that even though starting with something can save you time, it can also cost you a ton of time so it's frequently a wash or a net negative.

Personally I believe the stock market is incredibly, incredibly shaky. Investors are now in full-fear mode, it doesn't matter what news Nvidia etc print - if customers of OAI and others, are not seeing a meaningful INCREMENTAL increase in revenue generation or increase in cost-reduction (aside from white-washing it with lay-offs from insane hiring in the past).

RE. stupidity - it is stupidity for the most part. Without the stupidity in quantity of demand, there is no market for LLMs from enterprise et al.

Wanna know how stupid it is? Someone I know who works at Blackrock as a portfolio manager pretty high up is all of a sudden being forced to learn how to code and use LLMs to code. Yes you heard me right - this behaviour is expanding out of the software engineering profession.

Its absolutely nuts.

Right, so they applied to a couple of jobs and it worked for them?

I'm sorry, do you understand how uncommon and rare that is? sure, if their domain was REALLY niche and the jobs weren't publicly advertised, then i could see how that would work. but the experience is VASTLY different outside such niche cases

To be blunt about it, there's a decent chance I'll be quitting this job later this year, largely because of the AI push. I just hate these tools and I do not want to work this way. Losing an employee is a pretty big cost to the company. I guess the AI stuff is probably worth it to them, but there's a downside to it, too.

However, besides a few trades that use unions/licensure/apprenticeship as an artificial supply limit, most trades are only limited by a willingness to do the work. A few decades ago, trade work was much less expensive, because supply was higher and many did their own DIY, which limited what prices the market would tolerate.

If a stat like that is not accurately measured, it's useless.

That’s gambling. You’re truncating the curve below the top. It’s a terrific strategy for middlemen. Its expected value is lower than index investing.

My experience concerning using "Mathematica" instead of "Wolfram" in AI tasks is similar.

There is no point at which this argument will not be made. Therefore, it is a useless argument.

to tie back to the actually article, if you believe a rug pull is imminent then you got to get off the rug. Idk, you have to make a decision because we're certainly at a fork in the road. There's no guarantee waiting will result in a better outcome nor one saying it will be a worse outcome. There's going to be winners and losers always and lot of it is really just luck in timing. I guess, in reality, the careers we've built come down to a flip of a coin; stay on the rug, get off the rug.

/i'm thinking of buying a welding truck and getting in to that, then hire a welder and rinse repeat until i have a welding business. There's plenty of pipe fence in my neck of the woods and i see "welder wanted" all over the place so there's opportuntiy too.

It will put and end to the middle class entirely, but that’s the intent.

The reality is a lot of people who were formerly middle or upper middle class, and even some lower class populations will face steep, irreversible “status adjustment”.

I’m not talking about “we used to be able to take vacations and now we can’t”. I’m talking about “we used to be highly paid professionals now we’re viciously competing for low paid day labor (gig work) to hopefully be able to afford the cheap cuts this week”.

Absolutely not, a lot was done just because it was pushed as the current fashion and advertised to be solving problems that either weren’t applicable to the concrete use case or that it didn’t actually solve.

If there were a way to be a true Robin Hood and only extract wealth from the wealthy and redistribute that to poor, I'd call that a noble cause, although finance is not my field (nor is crime, for that matter) so it's not for me.

My chosen wealth multiplier is working at a community-owned cooperative, building the wealth for others directly.

false. The article is from 4th of March 2026, less than a month ago.

Also known as PTSD-induced amnesia, haha. We all tried to forget.

Nintendo also has similar vibes. I see shareholder calls asking about AI usage and their answers come down to something like "we're not ruling it out, but we'll only use it when a situation presents itself". They tend to be pretty good at pushing back against their shareholders. Having a proper war chest instead of constantly funding on debt probably helps.

> it is stupidity for the most part. Without the stupidity in quantity of demand, there is no market for LLMs from enterprise et al.

Stupidity implies incompetence and lack of intent. Greed is incredibly intentional. There's always a bit of stupidity with greed (we even call such an algorithmic approach the "greedy method" after all), but I think they are important distinctions.

I'll admit your blackrock example is plain stupidity, though. I know part of the end-goal is for "idea guys" to be able to make their ideas without pesky employees, but I don't think too many really think they can achieve that today.

Car washes are automated even though they haven't answered the edge cases of how to wash your car when your car is rolled on its side or a terrorist is actively blowing up the equipment. They simply only operate when your car is right side up (and other conditions, like in neutral, wipers off, and a driver who is willing to not exit the vehicle) and when there aren't active bombings on the building. And other "edge" cases.

Just because there is a possibility for something to not work, doesn't make it useless. Automated tire replacements could start with very rigid cases where they are applicable, and expact the scope slowly to allow more cases, like a bent wheel or poor weather.

I hope everything works out well for you.

How HTML, CSS and javascript come together is extremely relevant to developers 20 years ago and today.

I do support and agree with the parent comment, see the discussion, but I do credit getting into web development when it was raw and open paid dividends for me. Todays ecosystem is opaque in comparison. You don't think there's more friction today?

So then Claude starts discecting the instructions. I start writing some code.

After a while Claude is done, and I've written about two or three dozen lines of code. Claude is way off, so I have to think about why and then write more instructions for it to follow. Then I continue coding.

After a while Claude is done, and I've written about three dozen more lines of code. Claude is closer this time, but still not right. Round 3 of thinking about how Claude got it wrong and what to tell it to do now. Then I continue coding.

After a while Claude is done (yet again), and I've written a lot more code and tested it and it's working as needed. The output Claude came up with is just a little bit off, so I have it rework the output a little bit and tell it to run again.

I downloaded the resulting code Claude wrote and compared it to my solution, and I will take my solution every single time. Claude wrote a bloated monstrosity.

This is my experience with "AI", and I'm honestly not loving it.

It does sometimes save me time converting code from one language to another (when it works), or implementing simple things based on existing code (when it works), and a few other tasks (when it works), but overall I end up asking myself over and over "Is this really how developers want the future to be?"

I'm skeptical that these LLM-based coding tools will ever get good enough to not make me feel ill about wasting my time typing instructions to them to produce code that is bloated and mostly not reusable.

So I'm extremely bitter about this potential direction

AI isn’t like this because the final output is the same as hand coding.

I wouldn't "invest" in lottery tickets because for these p is far too small (exception: if I found a loophole in the system of the lottery, which has been found for some lotteries). For casinos, there is additionally the very important aspect that the casino will scam you (if you start winning money (for example by having found some clever strategy that gives you an advantage), the security will escort you out of the building and ban you from entering the casino again).

So, to give an explanation of the differences:

- Because "the typical run" for such an investment will be loosing, you should never invest your whole net worth (or a significant fraction thereof) into such an investment. The advice that I personally often give is to use index funds or stock investments for generating the money for investments that are much more risky, but have huge possible payouts.

- You should only do such an "early investment" if you have a significant information advantage over the average person. Such an advantage is plausible, for example, if you are deeply interested in technology topics

- Lottery tickets have an insanely small p (as defined in my comment). You only do "early investments" into topics where the p is still small, but not absurdly bad. The difference is that for lottery tickets the p is basically well-known. On the other hand, for "early investments", people can only estimate the p. Because of your information advantage from the previous point, you can estimate the p much better than other people, which gains you a strong advantage in picking the right "early investments" to choose.

But be aware that this is a strategy for risk-affine people. If you aren't, you better stay, for example, with index funds.

Do you have an empirical study to support that your employer should buy you a laptop and possibly a monitor or two to help your productivity?

If there's no study, why should we believe it?

It's like "A study found that parachutes were no more effective than empty backpacks at protecting jumpers from aircraft."

https://www.npr.org/sections/health-shots/2018/12/22/6790830...

And yes understanding them is still relevant. But when I started I was spending more time memorizing the the quirks of IE6 than I was learning how JavaScript, CSS, and HTML come together.

I think it you start directly in react you don’t learn the layer below it sure. But there’s no reason you have to start leaning react. There’s nothing inherent about starting today that forces you to start directly with React. You could start building a static webpage. And if you did that it would be easier and more fundamental than if you did that same thing 20 years ago because you can ignore most of the non-standard browser quirks.

And writing those instructions when I race it..it's more cognitive effort for me than coding!

If you’re paying a fair price for the risk, sure. Most of the examples you gave seemed to be in deep speculative territory to the point that they don’t very much resemble anything economic.

I also, candidly, haven't ever seen anyone successfully do that.

Are you under the impression that we don't bother to empirically prove things that seem obvious, like the safety benefits of parachutes? You don't think parachute manufacturers test their designs and quantify their performance?

I don't think that's making the argument you think it is.

Why can't you just pass any of those to an AI?

This is repeatedly used as an example in the medical community about the limits of randomized controlled trials. This isn't some impression - your impression that such evidence exists is wrong.

There might be some parachute company tests about effective of velocity, etc., but there are no human trials.

Why? Because that would be unethical.

When engineers demand evidence that AI is productive, but not that having laptops and monitors are productive, it screams confirmation bias. "I'm right, you're wrong" as a default prior.

It's a good thing "randomized controlled trials" aren't the only kind of empirical evidence, then.

We know the limits of how fast a human can safely land. Parachute manufactures have to prove that their designs meet the minimum performance specifications to achieve a safe speed. This proof is not invalidated by the fact that it doesn't include throwing some poor bastard with a placebo parachute out of an airplane to demonstrate that he dies on impact.

Also, the answer to your original question is yes. There are numerous studies showing that multiple monitors improve productivity.

I would emphasize that I don't think there's anything particularly wrong with the converse either. If an executive is just absolutely convinced that dual monitors are a scam and nobody needs more than their laptop screen, they can run their company that way, and I'm sure there are many successful companies with that philosophy.