Not to be the “ai” guy, but LLMs have helped me explore areas of human knowledge that I had postponed otherwise

I am of the age where the internet was pivotal to my education, but the teacher’s still said “don’t trust Wikipedia”

Said another way: I grew up on Google

I think many of us take free access to information for granted

With LLMs, we’ve essentially compressed humanity’s knowledge into a magic mirror

Depending on what you present to the mirror, you get some recombined reflection of the training set out

Is it perfect? No. Does it hallucinate? Yes. It it useful? Extremely.

As a kid that often struggled with questions he didn’t have the words for, Google was my salvation

It allowed me to search with words I did know, to learn about words I didn’t know

These new words both had answer and opened new questions

LLMs are like Google, but you can ask your exact question (and another)

Are they perfect? No.

The benefit of having expertise in some area, means I can see the limits of the technology.

LLMs are not great for novelty, and sometimes struggle with the state of the art (necessarily so).

Their biggest issue is when you walk blindly, LLMs will happily lead the unknowing junior astray.

But so will a blogpost about a new language, a new TS package with a bunch of stars on GitHub, or a new runtime that “simplifies devops”

The biggest tech from the last five years is undoubtedly the magic mirror

Whether it can evolve to Strong AI or not is yet to be seen (and I think unlikely!)

- Uv for Python

- Nix

- Performant Virtualization

- Ghostty

- DuckDB and, in general, performant OLAP

Don't get me wrong as I do feel the core of your thesis is correct. Emacs is my editor and I just finished writing a nicely recursive set of gMake for cloud a pipeline. Most of my core software tools haven't changed appreciably since the mid 2000s--right around the time git came out.

edit: I had no idea Nix was so old. I guess it just feels very "new" in my zeitgeist.

> all you could use was tech from 2000 and before you would have a pretty stable stack that just worked

The improvements made during the late 2000s and 2010s mostly had to do with making the functionality of these technologies accessible to non-technical users. I was younger and probably more mentally agile back then, but I remember the first iTouch I ever bought being very intuitive to use; you could usually intimate what you wanted to do without even looking it up. I got so accustomed to this intuitiveness (Windows Vista being an unhappy interruption in those series of memories) that by the time Windows 8 rolled around I was completely taken aback by how bad it was.

I mentioned in another comment that these productivity apps only really see a positive net expected value at the enterprise level, where they aren’t primarily used for efficiency but as coordination mechanisms and institutional memory. Individual users can only really hope to take advantage of them if they are intuitive to use.

From what I’ve observed, most of these UX failures are not the result of a lack of technical aptitude, nor an issue of cost, but of failures in institutional coordination (principal-agent problems and things like that) or the market simply being cornered; both follow the general trend of consolidation in the tech industry. The companies that are making most of our software are huge and they lack the competition to incentivize them to improve.

I installed some old Debian versions in virtual machines recently, and had a similar thought. Other than security upgrades really 99% of anything useful was already included ~25 years ago. Could probably go back quite a bit further. One annoying thing beyond ~20 years is going back to pre-UTF-8 and having to worry about 8-bit (sometimes 7-bit) character encodings, but that is the only obvious downside. Emacs versions around version 20 also were lacking things that I use today, but nothing that I could not learn to live without.

And you can install everything. As in, you can download (from their archive) the distribution ISOs from old Debian releases. For early version everything fits on a single DVD or single CD-ROM. That is thousands of libraries and applications. You don't have to think about disk space (or RAM) when installing things from there in 2025. Also everything runs very fast.

It's like hardware has finally caught up. The level of bloat from ~2000 is perfect for 2025, especially if you want to be able to set up and run virtual machines without worrying about resource use. For offline use running applications in virtual machines it is perfect.

I'd say Obsidian (just over five years old, since its first release), which is ironic because it's basically just a UI on top of text files.

No media other than plain text can come even close to the quantity and quality of tooling we have for it. Plain text is amazing to read, edit, share, version, tts, print and all of it with nearly maximum space efficientcy.

I feel like most things I use existed 5 years ago, but now they are just better versions of what they were 5 years ago. TypeScript, Rust, JetBrains IDEs, Firefox, Slack, iTerm2, Sublime Text, Apple iMessages.

For me:

- Obsidian notes with self hosted livesync

- VR <3

- 3D Printing

Probably a lot more that i can't think of right now. What I hate it cloud subscription services though

I hear you, but there are true moments of progress.

Vue is a huge improvement over jQuery, is the first one that roughly hit your timeframe.

> what technology in the past 5 years do you use [...]

I don't use any software made in the past 5 years.

I think software has improved in the last 20 years.

  - Linux container runtimes
  - Linux hardware support
  - NixOS (19.5 years old!)

My vi became vim became neovim. The keybindings are almost the same, but they adapt to newer virtual terminals.

As a programmer, my ability to express myself has got more nuanced. Programming languages have got better.

But the software itself doesn't seem to be better. Everything still depends on C, and the older programs live the longest.

Sqlite is better now than 20 years ago. Java is better now than Java 20 years ago. Linux init systems are better. Virtualization, containeristion etc are better.

WSL2, Neovim, LSPs, Brave Browser, fzf, yt-dlp - just the ones I've used today.

>>makefiles

They are hard to debug and I never could make the compilation as fast as with CMake (which sucks for many other reasons). Hopefully Zig build system will make both obsolete in the near future.

Actually I'll go against the grain here in saying this but I do find LLMs quite useful for a number of tasks. However you won't find an argument that the first two decades of the 21st century were mostly a waste of time in terms of what was built and how little the envelope got pushed outside machine learning. As an old backend developer I find the rise and fall of the nosql mania particularly infuriating.

Surely there are some!

I think AI is the obvious one. Also, VSCode (or whatever modern IDE you use) is definitely better than the IDEs that existed 20 years ago. LSP is fantastic. Hm... StackOverflow was definitely a step change over existing tools. Godot is really good, much better than anything that came before, IMO. Modern languages are pretty good these days - Rust and TypeScript are better than languages in the 2000s, to name two of the top of my head.