back
11 comments
I have every computing platform I've ever hacked on since 1978, sitting in my room in the form of a retro-computing collection - portions of which I have exhibited over the last few years, to the tune of a 40,000+ visitors.

As a systems analyst/programmer with 40+ years of experience, I have lived by the mantra that "technology doesn't get old - its users do", and that "every computer that was ever built is still just as useful as the very first moment it was ever powered on - all it needs is a user". These are radical maxims, its true - flying in the face of consumption culture - but it has resulted in another phenomenon, especially true now in the age of AI/ML: old computers can always become more useful through optimization.

A case in point, one of my favourite systems is the Oric-1/Atmos computer - an also-ran 8-bit system from the heigh days of personal computing, before the IBM PC took off. It is now possible, with AI/ML, to write extremely optimized software for this platform which really push its limits into new, creative territory.

As AI/ML make proton-/electron-basedknowledge less intangible, the actual atoms are useful again.

It is the physicality of the device, not the knowledge about it, which matters now.

I operate on an extremely elevated awareness of the idea of optimizing the bloat out of these older systems, using AI/ML tooling. Things are getting less and less sub-optimal, in the 'older technology hardware' department - we are able to really optimize software for these systems, now.

I think anyone with an interest in computing, should have a few retro computers in their collection - and they should be running modern software on it. Old computers do not die - their users do. There will always be a user for a computer, old or new. The distinction is entirely arbitrary, human.

Where's the utility? If it's a fun hobby to optimise old hardware that's fine. But there's nothing noble about using AI and expensive modern GPUs to get 5% more juice out of old hardware.
In my collection I have an Oric Atmos which was used for 40 years as a membership and fuel-tracking terminal for a motorcycle club. It was in this capacity, extremely useful to 250 members of the club, who would ride their motorbikes out into the countryside, unlock the clubhouse, type in their name, their fuel use, their mileage in a few simple keystrokes, turn it off, lock up the clubhouse again, and go on their merry way. They did this for 40 years before someone suggested they replace it with an Android tablet, with a web page, and all the latest bells and whistles.

It was donated to my collection after the tablet was integrated .. but of course, the membership and fuel-tracking capabilities of the club are now radically degraded, nobody can get the web page working any more, no data is being recorded by the club, and the whole thing was a fiasco.

That Atmos is still quite useful for that function. It still works, just as reliably as it has, for 40 years. The only reason it was taken out of service was because the consumer-/consumption- culture was applied to it - it was not 'new', it was 'old', it therefore needed to be replaced with 'something modern'. Indeed, if the utility question was asked properly - as you have done - the machine would still be there, functioning just as well as it had for decades.

Its not a hobby. Its a bet that, in the future, 'old computers' will find new uses. Every single one of the systems in my collection can find new users. The users just have to discard the consumer-/consumption neurosis which afflicts everyone ruled by the "new bar of soap" conundrum.

>But there's nothing noble about using AI and expensive modern GPUs to get 5% more juice out of old hardware.

I disagree with you completely, but then I focus on the user, not the use. There's a distinct difference. It is entirely noble to use a computing system for which there is absolutely no possibility that it be subverted for displaying ads or piping the data through an unknown third-party entity, for example, to maintain a private membership database and associated data logging ..

I’ve just been trying to get my hands on local compute period. I have this (perhaps irrational l) fear that personal computers will be increasingly expensive for the next half decade or so due to AI demand and I value having my own device much more now than just a chat bot.
Except the vintage computing space is already full of vibe-coded "retro" projects, so if escape from AI hype is what you're after, you won't find any refuge there.
On the other hand, the retro- computing scene is a perfect example of how AI/ML can really, really improve things - and anyone being myopic about it would get a serious kick in the ass if they take an honest look at the improvements being made...
I enjoy retrocomputing a lot, being a huge part of my life.

I do not want to see advancements made by a technology I truly despise. I'd rather see less impressive, handmade projects (therefore, more impressive) rather than slop.

Feel free to kick my ass.

AI/ML myopia is about as useful as Linux-myopia was in the 90's. Its here, its not going anywhere, it is getting widespread adoption.

Slop is real, but then so are accurate, timely, and well delivered products. Its up to the human, not the tool - same as it ever was.

You've made your decision. You're kicking your own ass.

>AI/ML myopia is about as useful as Linux-myopia was in the 90's.

Not only is this argument a copycat of cryptobro discourse, but considering that the first truly usable Linux release was in 1999 and no one expected what was bound to happen with Solaris, this is bollocks at-best.

>Its here, its not going anywhere, it is getting widespread adoption.

Getting shoved up in everyone's lives does not mean adoption. Nobody makes money out of this (except the Nvidia ponzi) yet, so that's a bold assumption.

>accurate, timely, and well delivered products

Have yet to see one 100% vibe-slopped thing that features these criteria and isn't an abomination of maintenance.

>You've made your decision. You're kicking your own ass.

And proud of it! ;)

The first really usable Linux release was way before 1999.

"cryptobro" is just a useless, troglodytic slur.

> Nobody makes money out of this

Not true, you can make money with any new technology if you apply your tool-using hat and dissuade the inner cannibal rage.

>Have yet to see one 100% vibe-slopped thing that features these criteria and isn't an abomination of maintenance.

Thats on you, homie. You won't see what you can't confront.

So stay away from parts and software made after 2024, shouldn't be too hard if 1990s era msdos is what you're after...