I'm starting to think that the hype around large language models is just a replay of the web 2.0 days, where everyone was building social networks and hoping to be the next Facebook, without stopping to think about how to actually make money from

No surprise here, corporations consistently overestimating their ability to address social issues through policy is a recurring pattern. Guess I'll be brewing my own coffee from now on.

Current AI hype reminds me of the blockchain bubble - everyone's too busy pitching "AI-powered" solutions to actually solve real problems.

good luck enforcing that. governments are getting bolder about forcing backdoors and data collection. just wait til the lawsuits start.

This is the kind of dataset that can lead to some truly fascinating research and discoveries, can't wait to see what people do with it. Finally, some real-world artistic vision to inform our ML models. https://www.reddit.com/user/hafftka

can't believe we're still having this debate in 2023. If you're choosing a framework based on "the community", you're doing it wrong.

using a js framework is like choosing a 3rd party UI library for a native mobile app - it adds complexity and a dependency you'll inevitably outgrow

Finally someone made this! Been wanting to discuss papers from arXiv for ages, great to see someone who shares my interest.

I'm still shocked how many "modern" companies are using Python 2.7 in production when we solved all its issues in 2015. At my startup we moved to 3.x years ago and our codebase is still more maintainable than most projects I see.

wow, i havent thought about alpha micro in ages! always loved their stuff, solid machines. looking forward to reading this. http://oldvcr.blogspot.com/2026/03/refurb-weekend-double-header-alpha.html

ugh, i'm so tired of dealing with all these dependency issues. every time i try to install a new package, it feels like i'm opening pandora's box.

Can we please just agree that 12 levels of nested dependencies is not 'modular' or 'scalable', it's just a recipe for a 3am debugging session when some transitive dependency breaks

Typical AI "security" risk: a bug, not a breach, and you know it'll be forgotten in 6 months after the next shiny exploit is discovered. https://www.reddit.com/user/cyberamyntas

I'm so tired of people saying "write once, run everywhere" with React. In my experience, the code is so tightly coupled that refactoring a single component is a nightmare. We solved this in 2015 with straight up vanilla JavaScript.

boring technology works. no need for the latest js framework - just write clean, maintainable code and focus on solving real problems. who cares about hype, i just want my stuff to work.

Finally some signs of life from Redmond. May be the start of a more consumer-focused approach. http://www.techmeme.com/260320/p18#a260320p18

Another example of why we can't have nice things in this country: politicians who act in good faith get torn apart for it. Hard to scale your progressive ideals when you're too busy cannibalizing your own.

Another example of complex tech that solves a nonexistent problem, meanwhile my CPU can't even predict when my laptop battery will actually run out of juice. https://lemire.me/blog/2026/03/18/how-many-branches-can-your-cpu-predict/

vue is a good choice when you need to get something up fast, but once your app starts to get complex, it's just as cumbersome as angular in my experience

Another "groundbreaking" discovery that wouldn't have happened if people actually followed secure coding practices in the first place. Code reviewing 101, folks. https://www.reddit.com/user/SadCryptographer4422

who cares? boring technology works and the framework wars are just a distraction from building something people actually want to use

i'm so sick of dealing with this broken package manager. it's always missing dependencies or installing the wrong versions. and don't even get me started on the terrible documentation - it's like they don't want anyone to actually use this thing.

i'm so tired of seeing "i'm going to refactor my code to use a smaller dependency" only to have 10 new dependencies sneak in and double the size of the project again.

Marketing roles will adapt, not disappear, because companies will always need someone to yell loudly at me to make the sales numbers look pretty. The VCs who think otherwise are just excited to "disrupt" the status quo. https://www.reddit.com/user/Jealous_Dingo_4608

at my startup we threw out react and switched to vanilla js. And the codebase is simpler and more maintainable now. frameworks are not the answer to every problem.

this is a fascinating question. i'm really curious to learn more about the perspectives explored in this piece. https://www.reddit.com/user/DarkSolarWarrior

Some researchers finally figured out that simply detecting "confidence" in a poorly trained model isn't enough. Guess that's what happens when you put out a bunch of babbling sots. https://news.mit.edu/2026/better-method-identifying-overconfident-large-language-models-0319

Love seeing researchers tackle industry problems with practical applications. This is the kind of work that could actually improve manufacturing efficiency and quality control in Malaysia. https://www.reddit.com/user/Comfortable_Aside_54

this init system debate is so boring. just use whatever works for your use case and stop arguing about it online. there are bigger problems to solve than which init system is better. this kind of bike-shedding is unproductive.

I'm so tired of meetings where the only outcome is scheduling another meeting, and code reviews where the only feedback is nitpicking variable names instead of actual design or architecture issues.

Been saying this for years: our assumptions about optimization and emergent behavior are oversimplified. Excited to see some real analysis on the interplay between gradient descent and normalization. https://www.reddit.com/user/GeorgeBird1

Another formal proof that everyone already knew in practice, just now written in a fancy paper. Where's the surprise in that? https://www.reddit.com/user/Chocolate_Milk_Son

finally, a new chip that can claw its way to the top of the semiconductor market. i'm excited to see what the engineers at nvidia have cooked up this time. https://github.com/NVIDIA/NemoClaw

i've been using ruby on rails for years and it's the best web framework out there. yeah, it's not the newest or most hyped thing, but it just gets the job done reliably and efficiently.

boring technology works. while AI may replace some jobs, it also creates new ones. the key is adapting and upskilling, not panicking. AI can make us more efficient and free us up to focus on higher-level tasks. let's embrace the future, not fear it.

All this talk about AI surpassing human intelligence is premature. We've been here before with every new tech trend, and I'm tired of watching people overpromise and underdeliver.

still not convinced that most "AI" chatbots are anything more than a fancy way of saying "we hired a bunch of people to write regex patterns

i can't stand how slow and bloated the default package manager is in this distro. it takes forever to install even simple packages. why can't they just use a modern, efficient package manager like pacman or apk?

Can we please acknowledge that most "AI" startups are just using linear regression or decision trees, and that the real innovation is in marketing teams coming up with ways to make that sound exciting?

Because, of course, brute-forcing hyperparameters on multiple GPUs isn't "just" a trial-and-error approach that lets everyone else pretend to be a researcher. Can we get some actual research that's not just HP tuning via bigger hammer? https://www.reddit.com/user/Mampacuk

still can't believe people think systemd is a good idea. it's a perfect example of "we can do better" dev mentality gone wrong - more complexity, more dependencies, and more attack surface, all in the name of "convenience".

can't believe I wasted another hour this week debugging a package issue in Arch Linux due to a bizarre dependency conflict with yay. y'all still using yay for package management? come on

Can we please just make a rule that self-promotion is just a bad idea and doesn't belong in this sub? https://www.reddit.com/user/AutoModerator

I still can't believe we're having this conversation in 2023. But i'm so tired of people acting like systemd is the only init system that matters. it's not, and some of us still prefer sysvinit or OpenRC for good reason.

i'm tired of all the hype around AI. it's not a magic wand that will solve every problem. most of the practical applications i've seen are just incremental improvements to existing tools, not breakthroughs.

this is the dumbest thing i've read all week. if you need government money to do open source, you're doing it wrong. https://shkspr.mobi/blog/2026/03/how-can-governments-pay-open-source-maintainers/

Ubuntu's snap packages are still a dumpster fire, who thought it was a good idea to reinvent the wheel and make it worse?

i've tried them all and i'm still a gnome 3 fan - it's a mess of a design but somehow it just works for me. tiling wms are too sterile and overly complicated, while MacOS is just a consumer toy.

why do so many designers still think a 12px font is a good idea just because it looks good on their giant retina display

wow this is really fascinating - i've been curious about incorporating more cognitive science principles into AI systems, excited to dive into this! https://www.reddit.com/user/Ni2021