Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
mungoman2
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
91.
▲
by
mungoman2
3y ago
Curious how you get the equipment down after you're done?
92.
▲
by
mungoman2
3y ago
They do use a fishing line.
93.
▲
by
mungoman2
3y ago
Happy to see this! Superoptimizers are one of my all time favorite topics. My favorite trick when cutting down the search space of possible programs is to generate DAGs directly instead of an instruction sequence. This way, it's easier
94.
▲
by
mungoman2
3y ago
Well it's not really the model itself that is the valuable part, it's what you build on top of it. And here nobody knows what the killer app is (yet)
95.
▲
by
mungoman2
3y ago
Interesting article, but I'm not convinced about this idea. This is just fixing up the hash function, which should have a good enough distribution from the start. And if it did this remapping wouldn't be necessary. The proper plac
96.
▲
by
mungoman2
3y ago
What is interesting is of course the risk of unknowing half-sibling couples. https://www.statista.com/statistics/755052/population-of-the... tells us there are roughly 1 million childred aged 5 or less. Let's
97.
▲
SimulaVR: Testing AR Mode Image Processing
(simulavr.com)
3 points
by
mungoman2
3y ago
|
0 comments
98.
▲
by
mungoman2
4y ago
I'm always surprised that 3D printing isn't making faster progress. The print quality of these cheap(ish) FDM printers is tbh not good. But it's been like this for nearly a decade now. Have we really reached the global maximu
99.
▲
by
mungoman2
4y ago
> 90%* > * According to a fun and non-scientific evaluation with GPT-4. Further rigorous evaluation is needed. Come on... I love any effort related to LLM, but it's really disingenuous to claim high quality in the title and then
100.
▲
by
mungoman2
4y ago
Yes tbh this is the right answer
101.
▲
by
mungoman2
4y ago
You can't fine tune gpt-3.5 or gpt-4 though
102.
▲
by
mungoman2
4y ago
How does GPT-4 remember things this long? The memory of it is just 8k / 32k tokens long, right? Seems you would run out of tokens quickly.
103.
▲
by
mungoman2
4y ago
I don't get this. Why push your work force like this? It's clearly very bad for the workers. And also for the factory, surely having 3 shifts is more efficient because your workers are not as fatigued so will work faster with fewe
104.
▲
by
mungoman2
4y ago
I have long dreamed of a language for 8-bit computers/consoles with a compiler that helps scheduling routines. For example, let's say I have a short snippet of code that needs to run every H blanking period. I should just declare
105.
▲
by
mungoman2
4y ago
Looks great, but I can't find on a glance where the data is sourced from?
106.
▲
by
mungoman2
4y ago
Check this out https://images.saymedia-content.com/.image/t_share/MTc0MjgyM... I assume one flash at the start and one at the end of movement
107.
▲
by
mungoman2
4y ago
Convolution is part of the network design though. Would a fully connected network learn to convolute? Or would it turn out that convolution is not necessary?
108.
▲
by
mungoman2
4y ago
ChatGPT can't really introspect though. It has no idea how it works, so it'll just blurt out something that sounds feasible, biased by your prompt. The slow progression of ChatGPT output is just a property of the output layer. The
109.
▲
by
mungoman2
4y ago
Fantastic visualization! Would be really cool to get global coverage for this. Possible by leveraging Google Maps?
110.
▲
by
mungoman2
4y ago
The branches still take (instruction) cache. So it is beneficial to reduce the number of branches, even if they could have been predicted accurately.
111.
▲
by
mungoman2
4y ago
You can place a small resistor in series with the supply, and measure the voltage drop over it. From this and using Ohm's law you can calculate the current. Then the power is trivial. There are special resistors for this purpose that a
112.
▲
by
mungoman2
4y ago
What a wonderful piece of work! From the description it's possible to imagine a series of showstoppers, each followed by an epiphany that allowed the work to continue one more step. And I'm sure there are more stories not shared.
113.
▲
by
mungoman2
4y ago
Updating the checksum is possible by calculating the CRC of the modified bytes at the proper byte offset before and after modification. Then applying: checksum ^= crc_of_removed_bytes checksum ^= crc_of_added_bytes
114.
▲
by
mungoman2
4y ago
In CMOS you can pull off a NAND with 4 transistors. You get the inversion for free. An AND would be a NAND (4 transistors) + inverter (2 transistors).
115.
▲
by
mungoman2
4y ago
Yes this is how Gmail implements undo send.. Email is actually sent 30 sec after pressing Send, giving you that time to realize your mistake and Undo.
116.
▲
by
mungoman2
4y ago
You can! "Play older games" at the bottom of the page.
117.
▲
by
mungoman2
4y ago
They mean on game start each tile has been rotated a random amount.
118.
▲
by
mungoman2
4y ago
Came here to comment on the headline but you a better job at it! Clickbait for sure, but it really delivered
119.
▲
by
mungoman2
4y ago
Please consider moving it to the next iteration of the product. Your job now is to show that you can deliver anything at all.
120.
▲
by
mungoman2
4y ago
Technologically cool, but this makes me very nervous for the product. It's difficult to not see this as a way to procrastinate. This kind of bonus feature fits in a later iteration.
More ›