Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
0x008
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
41 ms
·
511.
▲
by
0x008
6y ago
Yea you are absolutely right. While microservices do alleviate the aforementioned problems, sensible software design will as well.
512.
▲
by
0x008
6y ago
Remember when we said $800 for an iPhone is beyond expensive and unreasonable?
513.
▲
by
0x008
6y ago
If the mic on the AirPods max is anything like the AirPods Pro it will be hot garbage as well.
514.
▲
by
0x008
6y ago
Proprietary blue tooth implementation most likely?
515.
▲
by
0x008
6y ago
Bose launched new headphones at $400 and sell the old ones (3year old tech) at $180. So they will launch a „new improved“ line to compete at the $500 price bracket.
516.
▲
by
0x008
6y ago
And yet it is impossible to know all the requirements upfront which is why micro services and atomic functions are so popular. It seemingly reduces the complexity of redoing everything.
517.
▲
by
0x008
6y ago
“recovering”
518.
▲
by
0x008
6y ago
Also, besides memory speed which certainly will be an important factor, it might depend on the model architecture, fp32/fp16 etc. Much like you cannot say CPU 1 is faster than CPU. 2 - it totally depends on the type of workload - same
519.
▲
by
0x008
6y ago
Afaik the new tf alpha for m1 .Mac does not make use of the neural cores.
520.
▲
by
0x008
6y ago
Well I think we will find it roughly in the neighborhood of a 1080 (non-Ti) for optimized models.
521.
▲
by
0x008
6y ago
In computer vision you can go far wit 8GB VRAM.
522.
▲
by
0x008
6y ago
But that’s a CoreML model so that might not be possible with regular tensor flow models?
523.
▲
by
0x008
6y ago
I thought the tensor flow alpha for mac uses gpu hardware acceleration, not neural cores?
524.
▲
by
0x008
6y ago
I think the main point of the article is (and I would also emphasize this) to chose proven and widely used libraries - not languages. You want others to have experienced the pitfalls you might run into so you don’t have to wait or contribut
525.
▲
by
0x008
6y ago
I really appreciate the modernization of scala, but I fear the projects we have at work will never be migrated past 2.xx. We are just preparing the move to 2.12...
526.
▲
by
0x008
6y ago
The question is: can I save a password generator profile per entry? Because different sites sometimes require certain characters or forbid them. In Keepassxc it seems to be global which is not really useful.
527.
▲
by
0x008
6y ago
I tend to only write stuff down I learned from reading the book, then later turn these notes into flash cards. I feel the stuff that is obvious to me shouldn’t be written down, even if my notes then lack a certain completeness or are incomp
528.
▲
by
0x008
6y ago
Obligatory mention of foam for vs code here.
529.
▲
by
0x008
6y ago
Ah yes, the monthly note taking discussion.
530.
▲
by
0x008
6y ago
I’ve been reading the comments for an hour now and I have to say this discussion is amazing.
531.
▲
by
0x008
6y ago
Yes, the point I was trying to make that the large difference in value between tool and target that is implied is completely irrelevant.
532.
▲
by
0x008
6y ago
Dupe
533.
▲
by
0x008
6y ago
The only thing I don’t like about it is that I can’t figure out how to set a global hotkey on OS X.
534.
▲
by
0x008
6y ago
Strongbox is a nice iOS keepass compatible client. Integrates with the iOS password auto fill and can use can use Face ID / Touch ID with the pro versions. It supports google drive, iCloud, onedrive, sftp and local Storage on your iOS
535.
▲
by
0x008
6y ago
Much like you can hack a million dollar house with a sub-dollar lock pick?
536.
▲
by
0x008
6y ago
I have no Idea what this is for and what this is actually, but the website is beautiful.
537.
▲
by
0x008
6y ago
you should edit out the comment about female gender here.
538.
▲
by
0x008
6y ago
yes you can. But don't forget to estimate your power bill beforehand :)
539.
▲
by
0x008
6y ago
*back into a world
540.
▲
by
0x008
6y ago
And with rosetta it is certainly slower than running it native on x86
More ›