Over the years I have enjoyed many deep and interesting technical discussions online - both here on HN and elsewhere. I can't help feeling that those discussions have also all but disappeared over the past few months - alongside high quality free material that some experts have kindly given away but increasingly no longer do. I rarely find myself contributing any more because there is little on any of my usual online hangouts that is interesting enough to spark a deep technical discussion any more. It's 90% AI stuff - either written about AI or (usually poorly) written by AI - and most of the discussions that do initially catch my interest at all seem to end being very negative and depressing. (I appreciate the irony that I'm perpetuating that cycle right now. Looking through my relatively sparse recent contributions to HN is actually quite depressing. Although at least I actually am getting out more!)

I wouldn't mind so much if I believed the hype about the benefits of AI-generated work and thought that overall this period would be a net win for humanity. Unfortunately I don't really believe that. I fear the LLM era will eventually be remembered for producing tools that could do average/routine tasks quickly and to an acceptable standard but that were actually used to (try to) do good/innovative tasks with substandard results. But by the time enough decision-makers realise this it might be too late and the damage might already be done and hard to reverse.

But unlike the author of this post (or cited authors) I'm not sore. AI is providing a much better education model for students. So it's us that we have to adapt, not complain that the world is changing.

For example, in Machine Learning our projects are sometimes extremely long and repetitive. Because before AI, it was a valuable skill to be able to recall APIs quickly (what is the method to drop duplicates in a dataframe again?).

But today that's no longer the truth, and for the better. I am happy students don't have to go over the same repetitive low-value processes that I had to go to. Students can now focus on more high level valuable tasks: understand the domain, compare evaluation methods, compare models, balance tradeoffs, present solutions/reports, etc.

It's like in the 50s when students had to memorize tables of logarithms. Nobody misses that.

We are in the process of recreating all our projects to adapt to the new world. We'll see how it goes.

I use AI both in the cloud (I tend to avoid Anthropic...) and with local models.

I've probably written more code in the past year than in my whole career. I can now create what I desire (I just created a telemark skiing game over the weekend based on my Strava segments). I think this has helped my teaching as well. In a recent ML course, I created many interactive examples that in the past were just doodles on my whiteboard, but are now embedded in the notebooks I give students.

I've also taught a few AI courses to clients as well. However, with the inevitable rust/ASM-ification of everything, I'm not sure how long those skills will be valid. I do, however, think that having a human in the loop for ML and data analysis is still important.

I think our industry needs to transition to figure out what does it mean to build a quality system in the age of AI? It’s one thing to vibe code an application that works. But would you depend your livelihood on it? Would you get business funding and base the future of your team on it?

I think the challenge of the educational space is to figure out how to teach the judgement to build solid systems without relying on “hand coding”. For me I’ve been working on technology agnostic composable frameworks that help build that judgement. But honestly we are all just figuring this out together.

There will still be a need to be able to train system builders, the question really is what do they actually need to learn to be solid system builders in this AI era. And there’s a strong chance that pure programming language or tech stack learning may not be enough. This is what we educators need to figure out.

Superior to any teacher I've ever had, no matter how involved they are. I genuinely don't even see what my teacher is needed for beyond a requirement for me to be there. I can ask the bot "can you clarify X section" at 3am and it'll do in 1 second. In class, the teacher reads the slides, adds some extra points and takes questions. But my bot does all this, completely personalized to me.

Times move on, and I don't see teachers surviving. I'm hoping by next year the college will offer an "at home, self paced course" option. Theres no reason for me to be there.

That’s a shame. I’m surprised someone so well versed in web technologies has such a brittle setup. Cloudflare Pages, or if you you’d prefer to stay CDN-less, a $4 VPS, will handle any sort of traffic this blog can reasonably receive.

If I had a question about a new library now, I wouldn't google for a tutorial, I'd just ask ChatGPT.

Imagine: you have possibility to load desired SKILL into your brain/body - lets say it is fluency in foreign language so you can use it as native language. Will you choose it or "old way" which is hard, long and does not guarantee 100% mastering fluency? Or improving your body, making you faster, stronger with more endurance. Or SKILL allowing you breathe underwater. All of these and more are just human (of course not all humans) dreams. If you could choose, you really would choose old thing instead of new thing which makes life easier.

I guess we as humans should definitely think about making life easier and think at bigger scale - the Universe. We still do not know our oceans, lands and so on but imagine what could be done if humans were better - in all fields you can imagine. Integrating AI features is hard problem to solve but I hope we will find out solutions. Personally, I would like to obtain many skills easier way. We are still in times where we have to learn or at least read/verify content generated by computer models. But what if we had power of these models in our brains and bodies...

Web development, as much as some would like to think, is and has always been, complex.

I haven't tried learning Rust yet, and I'm sure as hell not going to lean on an LLM at all when I do.

I don't think what's happening is a uniquely AI phenomenon. If it wasn't AI, it was going to be something else. The ground is always shifting beneath you. It's happened to me in multiple fields through my career. You adapt or you pivot. Like a startup.

For many people who feel negatively affected by AI, my advice to them is to do what I did when my professional disciplines changed to something I didn't recognize: pivot. The great news is that AI makes it really easy to be a beginner in a new adjacent field.

This time my approach was better and so was the capability of the model. I can make a small change in one place and have the model propagate the change throughout the codebase. I'm still working directly with the code, and the agent doesn't have to make any important decisions about structure (which remain generally terrible). My familiarity with the codebase is the same as if I had made each edit myself.

It is definitely possible to build faster and better with an LLM and not give up control or the enjoyable part of coding. The energy in this space still seems to revolve around guiding the agent or letting it take over, the key is ignoring it entirely and discovering what works for you.

just because our old-brains did it "the hard way" and now we have these LLMs that can just make it happen, doesn't mean that our little ones want to/should skip the steps of understanding HOW IT WORKS

me and my kids have been making all sort of silly games and my oldest is starting to wonder HOW DOES THAT WORK, PAPA??? and i'm like, well, let me show you this...

a magical thing! what a time to be alive!!! there is joy, but you have to conjure it!!

I'd gently submit to the author that these are not opposites. Knowledge commodification is good, and so is deep understanding.

Good article.

But then around 2020 I realised kids actually have no clue how to use computers because their primary computer is a phone or a tablet. And there only experience with desktop computers are Chromebooks at school.

Now I suspect my generation will be the most knowledgable technical generation by far. The only issue for us is that no one will care.

As someone who used to write a lot of tech tutorials and has published a technical book, for me it's much less a financial thing and much more just sadness that something I spent a lot of my life doing, that used to generate emails every week from people saying thank you for helping them is now just completely worthless. I've spoken about this before here, but I feel like I've lost a large part of myself over the last couple of years because almost everything I used to do in my spare time now seems completely pointless.

LLMs with HTMX and plain old CSS and JavaScript (maybe Typescript) is the future. Getting back to little to no build times and frontends that don't take up 1+ GB of RAM on the clients

I pulled up the QR code menu website on my phone and grimaced.

“This is AI slop!”

My friends paused their scrolling. “What’s that?”

It is already here. That is the current reality.

Maybe when it comes to any coding-related learning, I can agree with that point.

However I think the future of education in general is very bleak. When we have plummeting scores for the PISA exam across the world [1], I fear people will get dumber and dumber.

There will always be those who thrive, but the easier it is to obtain information/knowledge, the less I think people have an incentive to do it. And I don't buy that AI will solve all intellectual jobs anyway.

[1] https://www.dw.com/en/pisa-2026-results-reading-math-scores/...

> But today that's no longer the truth, and for the better. I am happy students don't have to go over the same repetitive low-value processes that I had to go to. Students can now focus on more high level valuable tasks: understand the domain, compare evaluation methods, compare models, balance tradeoffs, present solutions/reports, etc.

You honestly believe that students that don't have the tenacity to stick with learning a simple set of verbs and grammar are going to have the tenacity to learn concepts steeped in jargon?

I do mean learn, which AI use has been repeatedly shown to be detrimental to.

It's a real shame a leader of an education company doesn't value the actual learning process. Those "low value" processes are the building blocks to understand and master high level concepts. I'm not spending my day to day thinking about the low level memory allocation in my job, but that knowledge and work helps immensely in how I approach the higher level assignments and wider architecture. And I still kick myself because my assembly knowledge is is only 1.5 semester's worth.

And yes, I still look up how to construct an array in C++. Not really because my memory is bad (okay, kind of), but more for the practice of reviewing API's and parameters and maybe making new revelations. I probably knew internally some 12 years ago I could set an initial capacity in one of the array's constructors. I didn't really "understand" the value in that until one random work day 4 years later as I was working through some in-house container packages.

You can even say that doesn't matter if you know what resize() is, but it's those little bits of learning nuggets that builds you up from a "fine engineer" to a "great engineer". Less the knowledge and more the work ethics built to obtain it.

How will skill building happen when LLMs and coding harnesses seemingly provide all the answers at one's fingertips?

Yes. I stopped looking at the code about 9 months ago. It's easier/better/quicker to learn how to get the LLM to audit the code.

Investors don't care if the code is LLM-generated, they're mostly assuming that now anyway. The VC's I've talked to recently are screening pitches using an LLM, they're embracing this tech.

If I recruited a new dev to the team now, I would expect them to be able to run multiple agentic coding instances simultaneously. I would not expect them to be hand-crafting code any more.

I had to go back to writing code to actually have it register.

Learning is friction. LLMs remove friction.

Have you independently verified your learning?

AI is moving fast, so I'm very open to the suggestion that Claude can do today what it couldn't six months ago. But my experience with folks who are learning from AI, historically, has been that they have high confidence in abilities they are not able to demonstrate when asked to show mastery through conventional problem solving.

Teachers and schools need to let LLMs do the things they are better at, and concentrate on having instructors do the things that LLMs cannot. Teachers that only read from the slides have always been a problem.

Your instructor should be having you dismantle an injector pump or bleed an air locked injector rail. You can read a book about how to do that, but you won’t really know how to do it until you’ve done it for reall

If we want an education system optimised for producing capable agent jockeys, then by all means this seems an efficient way to do so. But sometimes I wonder if, for all the modern wonders capitalism and individualism have brought us, we haven't lost track of being humans - all sharing a temporary seat on a fragile spaceship, hurtling through darkness and uncertainty.

Just keeping our heads to the AI grindstone until we die doesn't really seem to be a good answer to what to do with our existence.

You clearly have enough humility in your field to be able to soak up knowledge. That's great for you! I suggest your dedication is rare and teachers continue to be essential for society.

[1] https://www.joshwcomeau.com/css/anchor-positioning/

I think traditional web frontends might be dead for any applications with servers that can afford to keep server-side state of a logged-in user.

(Notably this includes almost any kind of AI application, because the LLM costs dwarf the web hosting costs anyway.)

And for the stateless ones, there's always htmx

1. LLM webpages in my experience have definitely not gotten simpler, nor load faster. It just pushes out more of the same junk underneath without any of the oversight we'd give a human

2. I'm not even sure the webpages are the big resource hog anymore. Chromium itself has had the reputation for eating RAM for well over a decade at this point.

3. In many ways, we did this to ourselves. In the name of shipping faster, we (or perhaps the executives) outsourced solutions that would sacrifice the simplicity of the code for "more mainainable, cross platform" frameworks. Electron was a solution for software companies not wanting a dedicated Windows and Mac (and perhaps Linux) team. React was a very specific solution for one of the largest websites in the world, that the rest of the industry adopted as a trend.