davydm

thanks - i can't even get to the paywall because of the ai-driven enshittification of the web - i was blocked before I even got there.

That's a pretty positive way to say that the AI grifters have effectively enshittified the net. Every page is behind a bot detection interstitial, wasting users' time to get to what they're looking for. And when that bot detection is broken (and it has been, recently, a few times), then most of the internet is just plain broken.

I honestly wish the whole AI lot would go up in flames. Sooner rather than later.

Webpack is really powerful, however, it's a huge generalist, and I think the specializing tools are showing their speed. In addition, webpack has suffered the same unnecessary churn that a lot of J's projects do, where upgrades become incredibly difficult due to arbitrary changes without (easily implementable) backwards compatibility. The cheese moved, but to little or no benefit for the consumer. Ask me - I've been trying to upgrade a vue2/webpack4 project to vue3/webpack5 for a long time but can never get far enough - always hit a wall. I've been looking at ways to move to vite specifically because of this, and I thought I was blocked by some rather deep integration into webpack which facilitates building branded versions of our white label app, but I think I see a way with vite.

Webpack honestly needs official, guided tooling configurators. Documentation often mentions a block of code, but not exactly _where_ to put it. AI agents are apparently stumped by a 4-5 upgrade, documentation lets me down, no automated upgrade tooling, and a lot of the changes I've seen are just cheese movement - add nothing useful, but require upgrade maintenance.

If webpack wants to take the top spot again, they need to work on: - performance: this is the most obvious issue right now. Webpack builds aren't exactly fast, and the new breed of tooling, esp vite, blows webpack away - consistent api: stop moving cheese arbitrarily, or, if you have to change things, provide backwards compatibility shims or upgrade tooling - improvements to documentation: it should be so hard to figure out where suggested config blocks of code should go - providing real examples would help

To find out why, go read or listen to "everything is fucked" - optimism is the root of hope, and hope is what makes people buy into faith - religious, financial, social - all the "isms".

Hope is hollow but generally drives a lot of people forward. The trick is to either remain blissfully unaware of the mirage called hope, or find some way to live with full, honest, hopelessness.

I really like the "everything is a file" aspect of Linux. Windows does have the concept of mount points, but it's a little convoluted. Making that first-class with all volumes mounted off of root somewhere (even if down a few folders) makes life easier in a bunch of scenarios.

hold on thar pardner, it'll take a little longer for some of us :D

happy new year & +1

poohbear needs to fuck right off

How unsurprising.

Everybody trying so hard to offload ai on people, but the good news is people are wising up and rejecting the enshittification. $8m is tiny next to the ms-nvidia-openai circle jerk, so at least she shouldn't be totally broken when it all finally folds in on itself.

Practically anything made by an actual human who actually cares about what they're making.

Link seems dead, but the sentiment is spot-on.

"You're holding it wrong"

As long as I can turn it off, like I do now, and like they promise. But this is a little concerning.

yay, another vscode with ai integration

because, y'know, there's not enough of them

Simply put, the tools are not at the level the grifters want you to believe.

You'll always find someone claiming to have made a thing with ai alone, and some of these may even work to an extent, but the reality is that the models have zero understanding, so if you're looking to replicate something that already exists (or exists in well defined, open source parts), you're going to get further than if you're thinking truly outside the box.

Then there's library and framework churn. AI models aren't good with this (as evidenced by the hours I wanted trying to get any model to help me through a webpack4 to webpack5 upgrade. There was no retained context and no understanding, so it kept telling me to ado webpack4 things that don't work in 5).

So if you're going to make something that's easily replicated in a well-dpcunented framework with lots of stack overflow answers, you might get somewhere. Of course, you could have gotten then yourself with those same inputs, and, as a bonus, you'd actually understand what was done and be able to fix the inevitable issues, as well as extend with your own functionality. If you're asking for something more niche, you have to bring a lot more to the table, and you need a fantastic bullshit detector as the model will confidently lead you down the wrong path.

Simply put, ai is not the silver bullet it's sold as, and the lack of app explosion is just more evidence on that pile.

To be clear: it's our competitive nature that has fueled a lot of human progress and is responsible for where we are, but we're at a bit of a tipping point where the world is getting more crowded, competition is increasing, and tribalism is entrenched everywhere. We need a radical human transformation if we're going to ever break out of this cycle without annihilation.

nice idea - not exactly new - that we could just get AGI to do all the work and spend our lives playing

BUT

In the two major economic climates - capitalism and communism - this won't end up being the case. Capitalists will moan about supporting freebasers. Communists will moan about the people not being productive - it's not like fun, free time and personal choice are highly valued in a place like China where 996 is normal, and people simply don't have time for anything. No, that won't change - they're already "productive to the max" - I'm quite sure that if China could come up with a "relaxation drug" to mitigate burnout, it would be 997 or higher.

Before any of this utopian vision can be fulfilled, we have to overcome the greatest hurdle of all: people. People are selfish and like their "wealth ranking" when they aren't in the bottom pool. People want to feel special, elevated - and mostly, people achieve that by being "better than" someone else. A "Star Trek world" only exists once the participants can actually sit down and agree to co-operate instead of treating everything as a zero-sum game.

interesting

btw - can't connect to any consoletext.dev domain - dashboard, docs, etc

just more of the same "oh, if you think <insert name here> is bad, you should see our enemies"

nothing will change for as long as people continue to make excuses

ugh, and the ai apologetics - "you're holding it wrong"

no, it's garbage - if I have to earn a phd in machine-speak, I might as well continue on my rather deterministic tdd path instead of debugging more garbage aigen code.

"get ready for a wave of anti-AI populism"

buddy.

buddy.

it's long been here, and not just for the power-draw, but also for the absolute grifter garbage that it is - promises from the creators because they desperately need to recoup several countries' worth of gdp, and it doesn't really improve - we still have to reject the shit code, recovering meth addicts are still instructed to "treat themselves to a little meth", people are encouraged to kill themselves, fake news and bullshit aigen video and images made not for fun but to deceive - all of these issues and more are why current ai is really just an over-expensive toy that can't be trusted to be reliable in any way, but is sold as cognition to the silly sheep that keep believing the hype.

I personally can't wait for the bubble to pop. There are uses for "ai" as it is now, but most of those uses are not what it's being sold as being good at (no, sorry, your aigen code is still a greasy butt snickers), and the few arenas where it actually is kinda useful (eg in entertainment) it's either used for nefarious means ("entertainment by deception to get the numbers") or it's rejected by the users anyway (see recent "ai in games" stories).

awesome - lean into the privacy leaks

if this feature ever makes it to consumer-grade gpus (and why wouldn't it?) I hope someone smart figures out how to cripple/disable it.

Interesting to choose an immutable distro because "I never had a setup survive multiple years. At some point, package updates would always break something enough that it felt easier to just do a clean install rather than spending hours debugging." when I've never, not once, reloaded due to package manager issues. I've reloaded on the recommendation of the upstream team (Ubuntu release). I've reloaded to get a more up-to-date package index (eg debian stable -> Ubuntu). I've reloaded to get away from systemd and pulseaudio (ubuntu-based distro - forget which, but likely mint), and I've stayed on the same live rolling Gentoo machine without reloading for over a decade.

If your package manager is causing you to reload - find a different distro. And by that, I mean a completely different distro with a different package manager.

part of me says "wow, cool"

another part says "oof, reads like shit, sounds like shit, i really hope the authors don't mind that they flushed their lives in a toilet to make this crap" (yes, rude, i know)

how I wish people could stop this stupid {first letter}{count the letters}{last letter} naming - it just wastes people's time as they have to go look up what the abbreviation of the week is

No

1. AI still makes mistakes - you'll still need a mentor to guide it 2. AI doesn't learn - so all the companies leaning on AI to replace interns are just putting more pressure on whoever is required to clean up after the ai, but that pressure won't be relieved because the same mistakes will happen going forward.

These are just the immediate business reasons for the "no". The bigger over-arching reason is that if no-one is hired as a "junior", there's no-one ramping up from "just studied, full-confidence, zero experience" through "wow, I thought I knew stuff", "wow, I'm learning more and making fewer mistakes", "I'm capable enough to contribute to the team without hand-holding" and finally "I could mentor someone else". Replacing entry-level jobs with ai causes a proficiency vacuum down the road.

Of course, short-sightedness never stopped the greedy before.

alrighty, show it then...

I get your part about mentors. I came up through having to figure stuff out myself a lot via stack overflow and friends, where the biggest problem for me is usually how to ask the right question (eg with elastic search, having to find and understand "index" vs "store" - once I have those two terms, searching is a lot easier, and without them, it's a bit of a crapshoot). Mentors help here because they had to travel that road too and probably can translate from my description to the correct terms.

And I really wish I could trust an llm for that, or, indeed, any task. But I generally find answers fall into one of these useless buckets: 1. Reword the question as an answer (so common, so useless) 2. Trivial solutions that are correct - meaning one or two lines that are valid, but that I could have easily written myself quicker than getting an agent involved, and without the other detractors on this list 3. Wildly incorrect "solutions". I'm talking about code that doesn't even build because the llm can't take proper direction on which version of the library to refer to, so it keeps giving results based off old information that is no longer relevant. Try resolving a webpack 5 issue - you'll get a lot of webpack 4 answers and none of them will work, even if you specify webpack 5 4. The absolute worst: subtly incorrect solutions that seem correct and are confidently presented as correct. This has been my experience with basically every "oh wow, look what the llm can do" demo. I'm that annoying person who finds the big mid-demo.

The problems are: 1. A person inexperienced in the domain will flounder for ages trying out crap that doesn't work and understanding nothing of it. 2. A person experienced in the domain will spend a reasonable amount of time correcting the llm - and personally, I'd much rather write my own code via tdd-driven emergent design - I'll understand it, and it will be proven to work when it's done.

I see that proponents of the tech often gloss over this and don't realise that they're actually spending more time overall, especially when having to polish out all the bugs. Or maintain the system.

Use whatever you want, but I've got zero confidence in the models, and I prefer to write code instead of gambling. But to each, their own.

wtf

What a dumb piece, basically promoting consumerism in the name of "the economy". I think it's great that people are starting to wake up, waste less, and recognize the hollowness of perpetual upgrades.