> ...so it sounds like we're using net-smaller apps :)

I also remember circa 2000 when the CS meta was, "who cares about optimization when you're riding the perpetual tailcoat of Moore's law"...then homogeneous parallel architectures became a thing overnight and much of the software development community was caught with their pants down.

Flash forward over a decade later with low-power embedded being the dominant end user platform and heterogenous compute already at the competitive forefront. Perhaps it's personal bias, but when I read net-smaller apps, my hardware design instincts immediately kick and scream net-complacency.

My personal complaint is that, with 200x more compute, we should be doing 200x more and/or more powerful things. Instead we're doing the same things ~200x less efficiently.

Yep. If you look at many of the daily tools out there, they do basically the same as was done using CLI or TUI tools back in the day, but now they pile on graphics "assets" to make the experience shinier.

The last time I noticed a real performance increase was going from HDD to SSD. I mean, I do also ask a lot more of my machine than 10 or 15 years ago so it probably balances out.

What is the biggest current bottleneck? Going from HDD to SSD makes even a 5 year old laptop feel close to current. Is there something that would do the same now?

Past 10 years is getting the same performance with less energy so you don't have fan noise and you get long battery life. Past 5 years is soldering everything to motherboard so you can't upgrade or replace anything without buying a new laptop.

Depending on the CPU, RAM speed can be a big one. Laptop vs desktop still carries a performance cost due to the power requirements. NVMe SSDs are to regular SSDs what they were to hard drives, but there are limits, especially gigabit internet and slower CPUs. Honestly though, the biggest gains are generally in making software more efficient — like choosing Safari instead of Chrome, or removing unnecessary drivers and bloat from your system. Run a startup log or check your system’s logs for errors that might slow down your computer. Oh, and install browser-native adblockers on every device. :)

As blockers can be very bad for performance because they are dumb. uBlock Origin is good though.

Sadly, no. Clock speed and multi-thread aware apps haven't gone far enough to give that feel of 'snap' we all got. I remember my Intel x25m so fondly. Even dual nvme drives in raid0 don't make thing hugely faster. I'm interested in what one of the 18-core i9s would be like clocked at 5ghz with our optane storage. Probably 1/3 of a second faster to load Firefox....

M.2 SSD PCIe hard drives are amazing. You can boot Windows in under 15 seconds inside of a VirtualBox VM instance. So about the speed of what it would boot in if it were directly running off your SSD. PCIe SSDs can transfer data at speeds greater than regular SATA SSD drives.

With Spectre & Meltdown patches, looks like we took a step back.

To go further back into the 80s where we were hard locked at 64k for commodore or 640k for DOS Assembly became the norm and optimizing for everything was the common goal for the whole community. It ended up creating a hack centric environment but it sure wasn't A) pretty nor B) beginner friendly. I'm old and have fond memories of assembly on the C64 as a kid, but I think we are in a better place where we have higher level languages and where Racket (Lisp of your choice here) and Python can run on a micro controller.

I wish there was a community of actively-developed 2000-era software that ran on modern hardware. We'd have so much space...