What they bring in my mind that is deeply surprising is the realization that from such a simple objective function (predicting the next word) could emerge such a wide array of general capabilities. Before LLMs, neural networks accomplishing "narrow" tasks were understood to be powerful, but not in such a general way.
What would it mean for forms to be "solved"? They are solved if you want to stick to boring/vanilla technologies. They are not if you insist on solving them, once again, with React, and then again with React hooks, and again and again, ad infinitum.
It didn't die per se I would say, it mainly got subsumed into more general programming and problem solving patterns (search, heuristics, A*, that sort of thing). What probably "died" is the dream that those would be sufficient to achieve general intelligence.
I find some problems of AoC are sometimes on the thin line between "it's interesting and I might learn something new" vs "wasting my time with under-specified or ambiguous problem specs".. Part 2 of Day 1 lies dangerously in the latter, for me.
I'm pretty certain that I saw somewhere recently that Sebastian Hinton is in the same family tree as Geoffrey Hinton (deep learning pioneer), along with an impressive number of other illustrious people (George Boole!).
For a category of software as "boring" as an OS is, I really agree that "just working" is the ultimate, almost sacred quality. I simply can't stand anymore the incessant ritual of software updates, and even the rhythm of something cutting edge like VS Code, on my Ubuntu LTS, is slightly too frequent for my taste.
I'm not sure about the clashing philosophies, but useEffect is certainly a "nexus of difficulties" about learning and using React (with hooks), in terms of the mental model it requires, knowledge of not-so-obvious JS concepts like closures, etc..
I never thought of this, but now that I see your idea of the "two levels of state", I find it incredibly likely and clear that it's going to work like that eventually, yes.
When you think about it, adding "reddit" is a kind of counterpart mechanism to the rise of AI and generated content in a general sense. There's a real thirst for answers coming from real, genuine people, and adding "reddit" is the last resort we have to maximize the likelihood of it.
I have been using Linux as my primary OS for many years, but this particular issue is stubbornly baffling. There are solutions of course but they are not really good.
Many years ago I had the chance to visit the Pyramid of Giza but when it was time to go inside it, I've been instantly stopped by my visualization of how I would feel in this tight and claustrophobic space (almost crouching in some places, I've been told), especially if there are many people at the same time, and you cannot easily move.. So I'm very glad to be able to finally see it with such an amazing technology, thanks!
It's interesting to draw parallels between the way you describe it and the way more general large language models (LLMs, of which Copilot is in a sense, a specialized instance, applied to code, instead of general language) operate: they also always "know" how to answer any specific question, or how to complete any prompt, without any exception. A model which would be able to "show restraint", and "know when it doesn't know", would be a really impressive improvement to this technology in my opinion.
In particular, in this book there is a discussion about the dangers and epistemic distortions induced by the "supervised learning" paradigm in machine learning, which imposes rigid systems of classification on reality, which was a novel and refreshing angle for me.
I agree that the Go deployment model is generally a dream compared to Python. However, the other part of the equation is having access to mature and complete frameworks like Django (a high quality package blending best practices and common patterns), for which I'm not sure there's an equivalent (yet) in Go.
Many years ago I did some experimentation with the idea of slapping a very crude "parser" (no AI involved, just extremely basic string matching) on top of a Lone Wolf CYOA book which was made public via Project Aon [0]. I wrote about it here [1], and the result is still playable here [2].
They are talking about frameworks leveraging and extending the builtin Emacs completition mechanism [1]. The most well known ones are Ivy and Helm, I believe, but there are now many others.
Excellent video! While looking at the machine code listing from the C=64 German magazine from the 80s, I realized once more that there was an extra, ninth column for the hex code, which was quite upsetting for me as kid (I think it might even have slightly slowed down my learning of 6502 programming).. turns out it was a checksum code of course!
I recently watched a DistroTube episode [1] where he explains the many problems (cultural and technical) with the current Linux desktop environments, and he is very critical of Gnome, in particular. As an Ubuntu user with a very Emacs-centric workflow, it was slightly surprising to me, as I am quite happy with my Gnome setup, and I was not particularly aware of this "war" waging in the GUI dev world. However reading your comment I realize that I'm not alone in appreciating the extreme simplicity of Gnome, especially when it's almost entirely keyboard driven (I do everything I need by invoking the Super key). Caffeine is also the only extension I use, and my happiness would be truly complete with a keyboard shortcut for it, I think.
I have realized recently that a value that is gaining a lot of importance for me is the stability of my cognitive investments, in terms of tooling, across a time frame that goes beyond my "current project", and more into the "my entire career" territory. In that sense, I feel more and more that the choice of learning and owning Emacs, more than 20 years ago, while in school, has paid off tremendously, in the sense that it has always constituted a solid and consistent backbone from which I can always start a new project, no matter what the context is (corporate or personal). I feel this would have been harder to achieve with a tool which is asking me to update it every 3 days, and which periodically introduce important paradigm changes forcing me to reconsider the way I work. In short, I value Emacs' stability and consistency across time and tech fashions more than anything else, as I grow older.
I'd be also very interested to know more about your experience and ideas, and also, I would take the occasion to try to better understand what is exactly meant by "consultant", in this context (for some reason I would always want to distinguish between the sense where it's you working solo and independently for a company, and the other, where there's you, a company "hiring" you and their client).
To me the most intriguing ideas that have been explored by Wittgenstein are related to his notion of a "private language" (and the question of whether it is simply possible or meaningful). You can ask for instance what is the difference between your body pain, which seems to be a kind of "internal object", that only you can possess and access, and the blue of the sky, when you look at it, which doesn't feel like an internal sensation (rather like an external fact), even though it can be construed as one, if needed.
One interesting thing I noticed in the last few years is that the younger generation seems to have a modified version of your rule (1): they don't necessarily add the "er" part anymore (i.e. making the verb infinitive), which sounds (to me) grammatically strange and a bit confusing.
I'm a big fan of the compose key, both for its universality and intuitive "syntax". For many years I endured the pain of a dual French/US keyboard layout (one for writing, the other for coding), until a coworker showed me the special Option-key shortcuts you can use on a Mac to make French accents, and then finally I settled on the nice and intuitive Compose Key support offered natively by Ubuntu: https://help.ubuntu.com/community/ComposeKey
I really wanted to follow this class taught by Steven Pinker [1] recently, but the audio becomes so bad after the first lecture that it's almost impossible to follow. It made me also think that there could probably be some interesting post-processing applied, to improve the quality.
I believe this way of understanding AI and its implications is quite sound, and it reminds me of Rodney Brooks' Productivity Gain[0] article, where he argues that we should focus on talking about "digitalization", instead of trendier buzz terms like AI or whatever is in vogue at the moment (the toll taker's job on the highway has not been replaced by a sophisticated robot, but rather by a transponder, along with the digital networking backbone it must rely upon).