> It was actually just yesterday when I decided youtube shorts are no longer a fun way to kill time. There's a lot of amazing stuff to watch, but it's no fun any more, because anything you see that seems amazing is likely to be AI Generated which, for me, ruins it.
And the ones that aren't AI generated are badly clipped scenes from movies / TV shows with the same 5 royalty-free songs playing over them, which might as well have been produced by AI.
I've dabbled in microcontrollers and enjoy how the limitations force me to find creative solutions, but this is truly next level. I'm not great with a soldering iron but I'm seriously considering assembling one of these.
My wife teaches Algebra 1 and this is something she’s encountered with a few of her students. They’ve been able to get by more or less by using what we refer to as “advanced arithmetic” e.g. figuring out the variable in an equation by inspection rather than solving the equation. But of course that only gets you so far, and she’s worried they’ll bomb Algebra 2 next year. But no matter what she does they just can’t/won’t learn algebra.
Basically. The questions were easy enough that they were doing it in their heads. She figured it out when she made them show their work, and their "work" turned out to be just plugging the right number into the equation.
Huh... TIL I've been pronouncing it wrong. Either the Fresnel equations never came up while I was studying physics or my professor(s) also mispronounced it lol
In my experience, browser extensions are more one-and-done than network-level ad blocking like pi-hole. I can’t think of any time I’ve had to override a UBO rule but many popular pi-hole blocklists have a fair number of false positives.
> AI-assistance will be particularly helpful and powerful in large codebases (think 100k+ lines of code). It’s simply impossible for any human being to fully understand that amount of code in detail… Even PROs/craftsmen who enjoy writing the code manually can’t make such claims. Our context windows are simply not large enough.
This is a false premise because LLMs do not understand code. Yes, they may be capable of ingesting your entire massive codebase and spitting back out relevant stuff, but they don't actually understand any of it. Vibe coding is just "copy/paste from StackOverflow until it works" on steroids.
In The Pragmatic Programmer David Thomas and Andrew Hunt called it "Programming by Coincidence" (Chapter 6).
Programmers not actively thinking about their code might get it to work,
but they don't know why.
When it breaks,
the programmer doesn't know why and they can't fix it,
because they never knew why it worked it in the first place.
Refactoring a very large codebase into components that can be understood independently may seem daunting, but not doing it is a matter of unwillingness not incapability.
I believe LLMs are definitely able to, in some vague sense, understand the code they’re working with. If they’re able to produce meaningful output from the input, especially in abstract problems, they get the gist of what’s being coded, not just the tokens. That’s what they’re storing in their inner representations: higher-level view of the text they’re processing.
9% is reasonable. I've got pretty strict filters on my home DNS and it's currently blocking 12%. I imagine that number would be much higher if I didn't have ad block extensions on all my browsers and IoT devices on a restricted VLAN.
I've encountered fewer show-stopping bugs in Linux than macOS lately. And of the software that I use on both, the macOS versions have more problems. Honestly, the main thing holding me back from replacing my M1 MBA with a linux laptop is the wonderful speed and battery life. If the software problems get bad enough to negate those I'm switching.
And the ones that aren't AI generated are badly clipped scenes from movies / TV shows with the same 5 royalty-free songs playing over them, which might as well have been produced by AI.