There is no Software Progress

In fact software is going backwards by assuming it will be fast enough because of the new hardware, optimization is not needed anymore, which tends to make worse developers. Your audience is not representative of all devs, Electron apps, transpiled langs, etc.

Counterexample: new matrix multiplication algorithm. I don't know if we can call it software... but it's faster than it was 50 years ago. Discovering faster matrix multiplication algorithms with reinforcement learning | Nature

HTTP2 and to some degree HTTP3 / QUIC would be other examples of the network communication getting faster due to software

Check out what @KazeEmanuar is doing on the N64 with his rework of Mario 64 engine and what @yuzokoshiro did in the Sega Megadrive/Genesis with Earthion. It's very impressive.

I think you are right. And even in software there are constructs that were created long ago that take a long time to get fully realized. It would actually a Manhattan Project-like effort to try and create software that is fully "caught up" to the potential of existing hardware.

I get it but there is still progress I think, new languages being safer to use like rust could be considered a progress? But yeah, this is not the software golden age

the criteria seems bizarre. obviously, making use of the new hardware has much more priority than optimizing for an old one. also, it's a tight loop, a piece of hardware would not develop in its particular direction unless we knew that there're potential software gains waiting ahead. what you're talking about is lack of progress in hardware agnostic software which is usually not the most useful type of software.

There is donkeys writing ✍️ code. I include myself

Play multiplayer FPS games sitting in India under 30ms with servers in NA. That day I'll call progress.

I'll make it progress bitch, yes mf it's progressing

I suddenly got deeper understanding why @falco_girgis is doing what he is doing with gta:vc dreamcast port

Hardware is "easy" to improve because there are basically three dimensions you care about: compute speed, memory, and network bandwidth. Most software is not trying to specifically optimize anything

all of demoscene is about that, they regularly go beyond C64/AtariST/Gameboy/NES/SNES/<add your system here> limits in new demos

QUIC protocol?

Webrtc is pretty impressive that’s all been within the last ten years.

I think a good example of actual software progress is looking at C64 games in the early 80s vs games in the early 90s/the demo scene. But even then, a lot of that improvement comes from the improved developer experience of the PC. Also, floppy disks.

The progress in software was better when the hardware was too slow

And today we have teams spinning Kubernetes, CQRS, DDD, event stores for a CRUD Todo app. Progress ? what a joke !

Postgres?

Would what @VictorTaelin is doing with HVM and all that stuff constitute software progress? What do you think about it?

Sad truth 😓

@tsoding but thats a part of algorithm right, software is definitely progressing in a tiny way.

My dude you are using emacs. nvim can do things on my 2012 thinkpad which normal vim couldn't dream of on my gaming PC. My Linux desktop now has working audio and doesn't flicker when I record the screen.

I think that properly using modern hardware is an advancement as well, e.g. multi threading. I agree that almost no software cares.

That happens on a micro per project level. That’s optimization over time.

he is wrong

The Myth of Progress

It's because buying newer better hardware is cheaper and wayyy easier. stoop overthonk!