[richardw]

And these machines were originally built at a time when everything was simpler. We’re like space aliens arriving from the future, applying our advanced tech to build an old human relic. A couple motivated people can do it.

30 years from now: “HardwareGPT, make me an Amiga but with ZX Spectrum keys”

[galangalalgol]

I think there is an opportunity to use modern microcontrollers to created a similar simplified environment. The lack of a virtualizing mmu constrains ram sizes, which in turn constrains the degree of abstraction level bloat that otherwise occurs. Like living in a tiny house, everything must earn its continued place. That enables the understandability we remember. And a modern microcontroller has about the same memory as an amiga 1k while using a tiny fraction of the power for several times the processing capability.

I suppose we could just use contikiOS, it has a browser of sorts even. But I think the thread model that allows premptive threads has some footguns to adopt as a general desktop system, and a browser for these systems needs a more elegant way to handle web pages that are several times its ram size than just grabbing text.

[actionfromafar]

We are like space aliens from the future :)

Still, it's amazing to me that we can build stuff which is vaguely comparable to our current commercial capabilities.

Another similar observation... I loaded an LLM (local llama.cpp) on a 2015 laptop, almost a decade old, and had a coherent "conversation" with it.

If someone had time-travelled back in 2015 and started for me that very same program on the same laptop in 2015, it would have been confusing and downright scary. I would have started looking for network connections to the remote super computer which surely must be running this thing...

[qingcharles]

"Yes, the rubber ones!"

[vidarh]

"I can't do that. My instructions prohibit me from doing harm."