That sounds exactly like funneling money away from hiring and into AI investments to me.

As for COVID overhiring, Amazon doubled their workforce during the pandemic. I'm sure plenty of those hires are still employed. During their latest layoff, they also spoke about "shifting resources to invest more in artificial intelligence".[2]

[1] https://www.cnbc.com/2026/05/20/meta-layoffs-zuckerberg-says... [2] https://www.cnbc.com/2025/11/21/amazon-cut-thousands-of-engi...

* Economy is tighter overall.

* Covid overhiring followed by firing means fresh graduates face competition from a large cohort of people with a few years of work experience.

* Even if it's not necessarily true, if the C-suite _believes_ that AI can replace juniors, that's enough to funnel money away from hiring and into AI investments. When money for hiring is tight, seniors are prioritized.

* Like it or not, workforce immigration (such as H-1B) causes displacement.

* The number of CS graduates in the US has doubled over the last decade.

In Sweden, fresh graduates across a number of fields are having a hard time finding jobs. Many of thosee seem to have either a STEM education affected by a dip in "green" capex/investments, or in "overhead" sectors likely to be tightened in bad times, such as HR specialists.

Pretty much all of the stuff that was suggested back then or earlier: Shopping, advertising, video conferencing, collaboration, software distribution, media consumption, banking, finance and of course communication overall.

Most of these ideas weren't exactly new in 1997, but go back to services like CompuServe and even Douglas Engelbart's Mother of All Demos. The bottlenecks were bandwidth and personal computer performance (both of which were then predictably following Moore's law), not human imagination.

A few examples that a lot of people correctly extrapolated from: NLS (1968), PictureTel (1987) and later LiveShare, IndyCam (1993), CUSeeMee (1995), RealAudio (1995), RealVideo (1997).

Perhaps the core business problem with LLM:s isn't finding a product-market fit, but that our imaginations have been running wild with expectations on "AI" since at least the 1950s, and now we have something that quacks - but doesn't quite walk - like a duck.

I can see how an O2 can be a fun second machine at work however, as a conversation starter and mood lifter!

It's just that when I see someone running an SGI or equivalent these days, it's mostly Buttonfly or something to that effect. Maybe they're even running NetBSD, which seems even more pointless, since it gives the same exact experience you can have with any dirt cheap old PC. Is anyone still using them for, say, personal video editing or home project CAD drawings just for the heck of it? Or maybe solving Advent of Code?

I toyed a bit with a NeXT cube a while back. It was fun to tick that box on the workstation bingo sheet, but the excitement wore off rather fast; running an old version of Mathematica very slowly isn't my personal idea of fun. Similarly, I tried a pair of SGI/CrystalEyes stereoscopic glasses together with an Indigo a while back, which felt like a fun novelty trinket and held my attention for about as long.

I'm glad there are enthusiasts around who care for these machines and keep them around for posterity, because I think they have great historical significance. I guess I'm just not into that particular flavor of retro computing.

SGI machines are extremely cool, but I don't quite grasp if collectors of old UNIX workstation use their machines regularly, and if so, what for.

Still, if I had the cash and desk space, I wouldn't mind a souped-up Indy that I could play around with for half an our once a year.

Comments URL: https://news.ycombinator.com/item?id=48077901

Points: 14

# Comments: 0

As TFA says, the problem is that accumulating knowledge takes time and effort, and the AI hype and expectations on LLM-assisted coding helps with rationalizing ever more short-sighted decisions that squander or hinder that process.

Everything today is a web app. If it doesn't exist and you want to vibe code it? It's probably going to become a web app, vibed using a web app.

The problem is, web apps are stupendous memory hogs. We're even seeing Chromebooks with 8 gigs of RAM now. LLM:s are all trained for and implemented in apps assuming the user can have $infinity browsers running, whether it's on their PC or on their phone. It's going to be very hard to change that in a way that's beneficial to what passes for business models at AI companies.

Ah, the paradoxes of modern software.

The current idiomatic way of doing web layouts was, back then, almost entirely theoretical. The reality was a cross-browser hell filled with onResize listeners, in turn calling code filled with browser-specific if statements. Entire JavaScript libraries were devoted to correctly identifying browsers in order for developers to take appropriate measures when writing UI code. Separate machines specifically devoted to running old versions of Internet Explorer had to be used during testing and development, in order to ensure end user compatibility.

In short: The web was not in any way, shape or form more convenient for developers than the RAD tools it replaced. But it was instant access multi-platform distribution which readily allowed for Cloud/SaaS subscription models.

Electron happened more as an afterthought, when the ease of distribution had already made web UIs, and hence web UI developers, hegemonic. Heck, even MS Office for the web predates React, Electron, and something as arcane as Internet Explorer 9.

Things have gotten much better, but we're still having to reinvent things that just existed natively in VB6 (DataGrid, anyone?) - and at the cost of increasingly complex toolchains and dependencies.

Yes, they were comfortable and easy to set up (and use), particularly when compared to web development.

> a platform where their best bet at dynamic layout is `OnResize()` and `SubmitButton.Enabled = False`

This is a great description of what web coding looked like for a very long time, _especially_ when it started replacing RAD tools like VB and Delphi. In fact, it still looks like this in many ways, except now you have a JSX property and React state for disabling the button, and a mess of complex tooling, setup and node modules just to get to that base level.

The web won not because of programmer convenience, but because it offered ease of distribution. Turns out everything else was secondary.

While I'm sure there are some wholly self-made virtuosos on the list, it does give off an air of apparent nepotism.

Artists couldn't apply for this, but were officially selected. The program was stopped in 2010, meaning no new recipients have been selected since. As far as I know, there's been no studies surrounding any measurable increase in artistic quality or artistic output.

It is of course easy to point out how deeply unfair such programs are on multiple levels. Unsurprisingly, many recipients have utilized loopholes in order to receive the grant despite having incomes and wealth well above the threshold.

Edit to clarify: Sweden still grants long-term stipends to various artists, sometimes up to a decade. What's described above is a guaranteed, life-long, basic income.

The guy who coded the actual Nohzdyve game (that runs on real ZX Spectrum hardware) is Matt Westcott aka Gasman. He's a demoscener and has made some brilliant speccie demos. https://demozoo.org/sceners/5879/

https://www.amigalove.com/viewtopic.php?t=1179

Applications could define their own as needed, of course (the pointer was just a hardware sprite).