I wanted to hook into the THREE object and explore the scene, but I wasn't able to figure out how to bring it back into scope after it's been optimized out of the js context, so instead I searched through the bundle to find where it unpacks the data and did that manually.
It just boggles the mind how you can simply write a 3D program with a ready-made library today, instantiate tens of thousands of objects in 3D space, and the whole thing will render in real time on a phone without you ever having to worry about how that incredible performance is possible.
The power of computers comes at least partly from the fact that for many practical problems, they let you effectively pretend that resource constraints don’t exist at all.
It does, it helps to stop and smell the roses occasionally and remember how far we’ve come.
My first proper computer (defined by programming on it) was a 3.5MHz single core processor with 48KB of RAM.
My current one is 16C/32T that can boost to 5.7GHz and has 64GB of RAM.
Considerably more than a million times the RAM and about a million times the processing power (if you factor clock speeds, core count, OOE, branch prediction, memory width and depending on workload etc).
I have more RAM in my house than every ZX Spectrum ever sold (about 5 million which comes out to ~240GB).
Adjusted for inflation a million spectrums (175 at 1982 prices) comes out to about 640 million quid.
My PC cost ~4000 in late 2022.
A million times faster for 0.000625% of the price, it’s been a hell of a ride.
My pet realization: most of the power increase has been spent on graphics at an ever growing resolution, including in resolution not only the number of pixels, that's a squared number, but also color depth and FPS.
Dealing with graphics has shadowed how hugely powerful the modern computers are. We're noticing now because of AI.
That said I think programmers do notice the performance outside of AI when they use software that surfaces how fast modern machines are which we do more than most.
I can remember when compiling a Linux kernel was “start it and go watch star gate” now I barely have time to boil the kettle for a cup of tea.
Definitely agree with the sentiment (also a Speccy 48K guy), but at the risk of being pedantic, I think you're double-counting: "million times faster" is for one Spectrum, the "0.000625% price" is for a million...
I'm unsure that the tiktok model works because it's designed around fast, easy to consume content, whereas scientific papers require sitting down and really digesting the material. It's much easier to read dense text on a desktop/tablet over mobile. The times where I read arxiv on mobile, it's really just the abstract. If you summarize each abstract into concise bullet points that might be quite useful.
Oh that's interesting. Almost like a project planner kind of thing?
I actually only recently tried Claude Code. I didn't realize it was such a big thing now. I also find the signal to noise ratio is really low when it comes to AI hype.
reply