timochka

I think that's more about telling users though that if they let an apl find local devices, that can be used to deduce your location.

Which is exactly the right amount.

Honestly, I hope the AI companies implode as well - and am fairly sure they will; OpenAI's entire business model is that they will be AI gatekeepers and everyone will have to pay them for access - but even now, the only way they can make this true is by artificially inflating the cost of inference hardware. That's simply not sustainable - cheap inference accelerators will come from China (ok, maybe somewhere else but at the moment US industrial policy is doing its best to make it China) and then it's game over for that business plan.

But while that makes OpenAI et al a shitty investment prospect, it won't stop AI taking your job if you don't adapt. You would be amazed at the capability of models that you can run locally right now. All the major providers can go out of business tomorrow, and the guy who is running Qwen locally will still be twice as effective a developer as you are.

The genie is absolutely out of the bottle and not going back in. The choice is to either find a new profession (which is totally valid, of course,) or start learning now how to be one of the people who survives by learning to use the tools better than anyone else. LLMs are dumb as a box of rocks, they are always going to need an actual intelligence to understand how to actually use them to accomplish any non-trivial useful thing. The people who get that and learn that will be the ones who still have a job when all this washes out.

(The "herp derp no they'll replace you too the only thing to do is wail and protest" contingent are inhabiting a weird space where they believe that LLMs are (a) useless but also (b) actually super intelligent and will replace us. Both these things cannot be true... Reality is they're very dumb, but still very useful. These things absolutely can be true; my compiler has zero intelligence, but it's still extremely useful and I'm glad I don't hand write assembler any more...)

He may be dead right, and unduly pessimistic at the same time.

The industry massively overhired relatively unskilled people who believed the job was "writing code", for a few decades; those people got by thanks to Stack Overflow and the fact that most of the code that needed writing really didn't need much skill anyway. There's very little innovation or engineering in most business platforms - a couple of decent architects and an army of code monkeys can deliver most of the software that's needed. We also developed programming languages and frameworks that made it much easier for the unskilled developers to be productive.

Sad to say, these are the people now loudly wailing about the iniquities of AI. If you thought churning out code was the profession - you absolutely should be very concerned that AI will replace you - it will. But if you realised that code was only ever a side effect of the job, which was applying computing to solve problems - you have little to worry about. AI is just a new tool to do it, and is no more a threat than optimising compilers, mamaged runtimes or IDEs were.

When I started my career, I wrote assembly code by hand for platforms that barely even had an operating system. Then C. Then Java - and as far as I'm concerned, every programming language since then has pretty much had training wheels permanently attached... For the last few years, I've preferred Rust, although the job has been more about developing architectures for others to implement than coding myself. But my expectation is that I'll end my career writing instructions for AIs to implement; and that's absolutely fine. The job didn't change, just some of the intermediate representations I had to write to do it. The profession will probably go back to look more like the one I entered - fewer people doing it, but with more formal education in computer science, and much less "coding".

Which is not to say I don't think there's anything to worry about for the profession, there absolutely is. How the hell we identify the future senior engineers with the actual skill and aptitude required, when there are no junior roles to provide the necessary apprenticeship - and when the education system is struggling to adapt to AI use in class - is a massive and fundamental problem. That's what keeps me awake at night, not people who think writing CRUD APIs in C# is a divine gift.

You do realise that you're increasingly starting to sound like That Guy who rants "he uses a compiler to wrote his code! This is insane, it'll never compare to hand written assembly!", right?

Any remotely competent software engineer is using LLMs at least to find out what they're capable of. And yes, they are capable of many useful things. The reality of utility lies at a point between the extremes of "not at all" and "100% vibe coded".

You are welcome to do your hobby programming entirely with pen, paper and assembler - or just without any AI tools. It's a hobby, you do you. But seriously, leave the professionals alone.

Republicans.

LibreOffice still hasn't managed to sort out making the UI scalable on Wayland, so you're stuck with either needing a magnifying glass to see the icons or you have to stand in the next room over and nothing in between. Given how many years they've had to work on that, I feel like the odds of them developing a working web UI in my lifetime are pretty slim.

Word of advice - don't scavenge old /server/ hardware if you plan to put GPUs in - unless you really like heat and noise. Those machines from the likes of HP will take one look at a consumer graphics card in a PCI slot and decide they need to run all the fans at 100% to ward off evil spirits. Not to mention just the general ballache of proprietary PCI riser cards, PCI power cables, etc. etc.

You're definitely better off with taking someone's old gaming rig off their hands.

In terms of specs - value VRAM above everything else. A slow, old 3000 series card with 24GB of VRAM is much more useful than a brand new 5000 with 16GB. If you can find old RTX3090 24GB, they're kinda ideal.

The one thing I will say for modern cards though is that they're much better for power efficiency - and in particular idle power (which is important if you're running the thing always on.) For my main LLM machine I have two RTX5060Ti (32GB total), which at the time was the sweet spot for price/performance/power, and it's very nice that they idle around 3 or 4 watts. I bought them before the world went crazy and prices went mad though, so they may not be the sweet spot any more.

One you're in 32GB VRAM type territory, you can run really really good dense models like Qwen-3.6-27b at a decent quant, decent context size, and good performance for things like coding, or bigger MoE models for more general use (particularly then if you have good CPU and regular RAM for offloading to CPU. For use as an assistant (i.e. not an OpenClaw fully automated slop machine,) I use 3.6-27b as a daily driver in Claude Code, and basically never use Sonnet.

Tremendous work, have you considered a career on the stage? Sweeping it, perhaps?

For what it's worth, I had a pretty much identical experience a month or two back.

Plex woke up one day and decided that the TV in my living room and the server in my home-office were clearly so far apart that I'd need to give them money to stream all 20 feet over my LAN - presumably because they woke up one morning and decided that it's more profitable not to understand VLANs (apparently not understanding VLANs is the "new Plex experience" and we should be very excited about it.) At least, that's what their support told me - they assured me that streaming from one room to another is now a paid feature.

Naturally, I told them to go fuck themselves and installed Jellyfin. And donated 10x what a 'Plex Pass' would have cost to the guy that made the Samsung-Tizen-Jellyfin-Installer thingummy. Because, well, fuck Plex.