Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
ynniv
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
91.
▲
by
ynniv
2y ago
The problem with evaluating LLMs is that there's a random component, and the specific wording of prompts is so important. I asked Claude to explain the problem, then write python to solve it. When it ran there was an exception, so I pa
92.
▲
by
ynniv
2y ago
The willingness of people to believe that combatants are wearing cinematic body cams for no tactical reason can only be matched by their willingness to assume people meticulously record every minute of their lives just so they can post a on
93.
▲
by
ynniv
2y ago
Well let me know when someone reproduces someone else's time traveling experiment. Until then I'm going to bet that the plethora of (untested, often philosophical) exotic QM interpretations are based on an incomplete model and are
94.
▲
by
ynniv
2y ago
Quantum mechanics seems intent on declaring that their particular map is the territory.
95.
▲
by
ynniv
2y ago
What the VC’s want is to say, “Good idea, kid. The grown-ups’ll take it from here.” But not this time. This is our time. This time you’re gonna hand ‘em a business card that says “I’m CEO...bitch”
96.
▲
by
ynniv
2y ago
It's dreaming Doom.
97.
▲
by
ynniv
2y ago
- Academia has become a paper throughput competition - Open access journals make it easier to game this system - Ergo, open access journals are bad Well, he's right that one of those things is a serious problem that needs t
98.
▲
by
ynniv
2y ago
You mean 21st century Apple. Because 20th century Apple had plenty of good flops: Newton, Pippin, Macintosh TV (no, not Apple TV). https://www.macrumors.com/2023/01/14/apples-biggest-hardware...
99.
▲
by
ynniv
2y ago
Looks like a JDAM with IR and radar final guidance. The attack is also interesting: instead of hitting the ship it aims a little to the side and explodes underwater: https://www.twz.com/new-look-at-air-forces-ship-killing-sm
100.
▲
by
ynniv
2y ago
GNUStep has never been popular, but it's out there (Aug 2023): https://www.mail-archive.com/discuss-gnustep@gnu.org/msg2550... As you guys know, Algoriddim makes use of GNUstep on Windows. It has just been releas
101.
▲
by
ynniv
2y ago
20+ years here. I've been exploring different ways of using them, and this is why I recommend that people start exploring now. IntelliJ with Copilot was an interesting start, but ended up being a bit of a toy. My early breakthrough was
102.
▲
by
ynniv
2y ago
Like Heptapod B. The "next word" argument is pervasive and disappointingly ridiculous. If you present an LLM with a logic puzzle and it gives you the correct answer, how did it "predict the next word"? Yes, the output ca
103.
▲
by
ynniv
2y ago
My disappointment comes from understanding that what humans do is keystroke prediction. If the output that I want can be solved by the most likely next keystroke, then sure, that’s a good use case. I’m perfectly capable of imagining those
104.
▲
by
ynniv
2y ago
The question is what does programming with an LLM get you over batteries-included frameworks with scaffolding like Rails or Django? Three years ago an LLM would conversationally describe what the code would look like. Two years ago it mig
105.
▲
by
ynniv
2y ago
My take requires a lot of salt, but… this time it’s different. Try writing single page web app or command line python app using the Claude 3.5 chat. Interact with it like you might in a pair programming session where you don’t have the keyb
106.
▲
by
ynniv
2y ago
The image presented is so distorted it makes everyday interpretation impossible. They'd be better off with a mesh of laser rangefinders and a camera shooting at a meager 1,000 fps. They say the winner was ahead by 5 thousandths anyway.
107.
▲
by
ynniv
2y ago
How long can a page of python take? https://github.com/Zyphra/tree_attention/blob/main/tree_shar...
108.
▲
by
ynniv
2y ago
If you're going to run kitty it can do a lot more than that: https://m.youtube.com/watch?v=ft1Q-DwGWIs https://notcurses.com/
109.
▲
by
ynniv
2y ago
He mentioned eventually running on the web, so my guess is to facilitate migrating old TUIs. There's plenty of consulting money in maintaining old enterprise software.
110.
▲
by
ynniv
2y ago
The landing page is refreshingly well done. It looks good, loads fast, feels smooth and stable. I also like the price point. It sounds like a lot for a saas, but doing it yourself takes too much time, and hiring someone is an order of magni
111.
▲
by
ynniv
2y ago
Except this happens with "real" intelligence in the same way. You didn't ask ChatGPT to decide whether or not it was just making things up, you asked it to answer your question.
112.
▲
by
ynniv
2y ago
"Asymmetry of nudges" is more directly conveyed as a ratchet: you can only change things in ways that benefit the corporation.
113.
▲
by
ynniv
2y ago
"Standard" doesn't imply "good", but that doesn't mean "non-standard" is better. Cost-per-quantity (L/100km) is easier to compare than quantity-per-cost (MPG) because how much you use isn't
114.
▲
by
ynniv
2y ago
The corporate world beckons you.
115.
▲
by
ynniv
2y ago
Why is playing Minecraft fun, but digging mines is not? Seems the same to me
116.
▲
by
ynniv
2y ago
lol, no. No it's not.
117.
▲
by
ynniv
2y ago
They're doing what they love.
118.
▲
by
ynniv
2y ago
See llamafile ( https://github.com/Mozilla-Ocho/llamafile ), a standalone packaging of llama.cpp that runs an LLM locally. It will use the GPU, but falls back on the CPU. CPU-only performance of small, quantized models i
119.
▲
by
ynniv
2y ago
I guess "decades of clicking things" is a domain one can be knowledgeable in? Usually boxes with draggable things on the top are inputs, on the bottom are outputs.
120.
▲
by
ynniv
2y ago
There are two outputs, two inputs, and three edges. The answer might not be obvious, but it's not a UI problem.
More ›