Wow! I am stunned how wrong that feels. I remember adopting git in the first year, and it still feels fairly recent. That it only took 10 years from Win95 to git, and 20 years from git to now, is truly uncanny. Win95 feels like a genuinely old thing and git like a fairly recent thing.
I don't know how old are you bit if you are in your 40s it's s just because you were a kid when Win95 came out and time seems longer when you are a kid (less routine, everything new, more attention all the time etc)
Probably a mix of this, and also that it feels like there have been many strikingly different milestones for Windows as a product since Win95, while git is not fundamentally different since 2006/07, or at least change has been far more gradual. Windows is so storied it feels older.
There’s been two main massive shifts that create before-and-after feelings in tech. One is going from “the computer is that super-typewriter that can send mail” to internet culture, and the second is going from online in pcs to always online in smartphones.
Win 95 feels from era1, xp and git was already in era 2.
Once those two changes were done by 2010 though, there’s been no game changer, if anything we've regressed through shittyfication (we seem to have fewer social networks vs the original Facebook for example, as most of them turned single player feed consumption).
Maybe pre and post LLMs will feel like an era change in a decade as well?
The hard part with 3d part creation isn’t the graphical interface or language, it’s actually describing and translating part requirements to a manufacturable design, weighing material, weight, fit, geometric, and cost tradeoffs. Openscad, opencascade, etc have been around for a long time and have specs for describing features in a way that llm should be able to handle, but if all the part constraints were available it’s far faster to make accurately in Solidworks.
This is my experience too. I took a course a long time ago in design for manufacturing, and it became abundantly clear that just because you can conceive of an idea doesn't mean that you can build it. That requires a lot more work and technical know-how that isn't always put into books or other "training data".
Just yesterday I had an LLM write an openscad module for generating a 2d rounded rectangle. It worked great! I then tried to get it to write a module to extrude a 2d shape into a 3d shape and it failed spectacularly several times before I gave up.
Interesting. I’m building a SaaS around this idea. And I managed to do things waaay more complex than that using LLMs. Especially “several times”. My AI can do a parametric trophy cup from one prompt in a couple of attempts, I would be shocked if it didn’t know how to make rectangular cube…
The Penn Station passenger display was, according to the NYT, segmented LCD glass made by Signature Technologies in Arizona.
It had 43 segments (each character had a 13 segment column, 17 segments column, then another 13 segment column that was a mirror of the first). You can see the segment shape on the original sign:
Step into a variant of a future, where claude is as important to the internet as aws, because constant interwebz rewrite happening in near real time had to be stopped for 4 hours causing incredible hacking spree as constant rewrites open/close/re-open various holes.
There is a part of me thinking that my initial thoughts on LLMs were not accurate ( like humanity's long term reaction to its impact ).
AI is going to make heroin look like a joke as more people integrate it into their lives. You're gonna have junkies doing some crazy shit just to get more AI credits.
I have a Frankenstein of a setup with this one. I use ZAI (GLM-4.6) for the base models, then Gemini's free tier for the search and image recognition. CCR intercepts the requests from Claude Code and sends them to each model/provider automatically.
I got annoyed at CCRs bloat and flakiness tho and replicated it in like 50 lines of Python. (Well, I asked Frankenstein to build it for me, obviously... what a time to be alive.)
I couldn't fix any of my UI quality-of-life bugs so I had to work on actual backend logic and distributed state consistency. Not what I wanted for an early morning coding sesh. Nightmare! /s
tl;dr Wayland doesn't have a good set of universally adopted input emulation and UI automation protocols yet, which makes a portable UI automation utility with the full scope of `xdotool` impossible to write. Work remains to be done to close this gap.
The X protocols in this area were not very good, but due to there being a single viable implementation you could rely on them being present (similar to using MSIE-only features in that browser's dominant era).
If you're German-speaking or can AI-translate the audio or subtitle tracks, I remember enjoying this 2021 documentary about the design process and project culture:
The GEM's super impressive, but a part of me also misses the charme and flair of the old Cairo Museum building, which had essentially become an artifact unto itself that somehow recalled that whole Howard Carter vibe we all subconsciously associate with Egyptology.
The very modern, flatly-lid, matter-of-fact galleries of the GEM I'm sure are SOTA museum work, but rather sterile. Still, I suppose the neutral looks have their virtue.
reply