Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
payne92
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
61.
▲
by
payne92
8y ago
Generally, yes -- they are retired. See: https://www.rhinocarhire.com/Car-Hire-Blog/July-2014/Retired...
62.
▲
by
payne92
8y ago
Title should be: “Xilinx acquires DeePhi Tech, a Chinese deep learning tech start up, for an undisclosed sum.”
63.
▲
by
payne92
8y ago
I wrote up a model for dividing equity in the very beginning, and it often ends up with different splits among co-founders. See: https://payne.org/blog/dividing-founder-equity-in-the-very-b... TL;DR: (a) figure out r
64.
▲
by
payne92
9y ago
For example, it's a little-known fact that Costco (in the US) sells caskets at pretty decent prices. See: https://www.costco.com/funeral.html Overnight delivery, shipping included.
65.
▲
by
payne92
9y ago
Facebook's biggest competitor is the US government. I'm not being flippant.
66.
▲
by
payne92
9y ago
Examples of questions about the existence of API and library functionality in the past four years? (with recommendation answers) Sure: https://stackoverflow.com/questions/46718944/facebook-graph-... https:/&
67.
▲
by
payne92
9y ago
I get what you're saying about their 'decision' to prohibit 'recommendation' questions, but my point is that it doesn't make any sense. It's a distinction without a difference, as many, many questions on
68.
▲
by
payne92
9y ago
SO is definitely a force, but I feel its usefulness has faded from inattention to the user experience and use cases. For example: #1. There's a huge "homework problem" (narrow, specific questions that aren't broadly us
69.
▲
by
payne92
9y ago
They're not "censoring people for discussing a perfectly legal technology"; read their policy carefully: >>> It’s important to note that this update to our policy does not prevent the discussion of related topics in
70.
▲
by
payne92
9y ago
> Most modern speech recognition systems rely on what is known as a Hidden Markov Model (HMM). This not correct. Most modern speech recognition systems are based on deep neural nets (DNN).
71.
▲
by
payne92
9y ago
It's very difficult to say: "productivity" has changed because the "products" that we work toward have changed. You spend an hour screwing around with a software problem while editing your video, but you're ed
72.
▲
by
payne92
9y ago
Geeky version: let's assume you want to "time release" something (e.g. unlock it in approximately 5 years). Encode your assets/keys/passwords and share just a portion of the key. If you extrapolate computing cost&#
73.
▲
by
payne92
9y ago
The boarding (lodging) house is back. We've come full cycle! https://en.wikipedia.org/wiki/Boarding_house
74.
▲
by
payne92
9y ago
Security is relative; there is no absolute security. SMS 2FA is not perfect, but it's MUCH better than what it replaced (nothing). ANY system with a human component will be at risk to social engineering attacks; not just SMS-based met
75.
▲
by
payne92
9y ago
Open always wins, given enough time.
76.
▲
by
payne92
9y ago
Prior discussions: https://news.ycombinator.com/item?id=14609525 https://news.ycombinator.com/item?id=16152050
77.
▲
by
payne92
9y ago
This is a terrible analysis from someone who offers professional investment advice (the author). Nobody would have held a 7.5% mortgage w/o refinancing. You can always ratchet down if rates go down. Mortgage interest is (was?) largely
78.
▲
by
payne92
9y ago
...through a cloud service. This is a critical caveat, IMHO. We are beginning an era were the largest cloud providers will attempt to differentiate with custom hardware. But the providers will keep the hardware to themselves, and just se
79.
▲
by
payne92
9y ago
There's an adage in poker that's applicable to crypto trading: if you don't know who the sucker is at the table....it's probably you.
80.
▲
by
payne92
9y ago
Virtually every industry has jobs like this, and many of those jobs are more physically demanding and much more dangerous. (And in some cases, pay much less). People seem happy to complain about warehouse working conditions and pay, yet gi
81.
▲
by
payne92
9y ago
"Win-win"...not at all. Patents like this have "threat value", which is often happily exploited by "IP monetization" companies, contingent law firms, etc. This is the kind of stuff that turns into 100x $50k set
82.
▲
by
payne92
9y ago
Here’s the issued patent: https://www.google.com/patents/US9817662 Note that it’s a claim on the processing unit implementation (e.g. the FPU), not the method. Nonetheless, I’d be very surprised if this stands the tes
83.
▲
by
payne92
9y ago
Very very cool. Boys and girls, this is how you do a resume. :)
84.
▲
by
payne92
9y ago
Disappointing, but not surprising. I've looked at quite a few drone startup projects (as investments). I'm not sure everyone fully understands how formidable DJI is right now. They are the Apple/Amazon AWS of the drone marke
85.
▲
by
payne92
9y ago
>> ...a careless approach to a main function can result in needless complexity. Compare two versions of the same main function: before and after. One version is over 400 lines. The other is about 40 lines. That’s an order of magnitud
86.
▲
by
payne92
9y ago
Oh Dear Lord. Programming IS abstraction. In fact, that's pretty much ALL programming is: using, defining, and implementing abstractions.
87.
▲
by
payne92
9y ago
From the complaint: Before any such “vulnerability” could have any chance to impact a user, the user would have to be subject to specific conditions and take the following steps: (1) the user would have to separately install the Keeper Br
88.
▲
by
payne92
9y ago
This article is off in so many dimensions. Fundamental research is not less active, but it's happening in different places (e.g. the Google Brain team). Find the most profitable companies and you'll find the research. And to sugg
89.
▲
by
payne92
9y ago
For the same reason people drive with their “check engine” light on: It’s frequently better to have a working system (i.e. “I’m late for work”), than to chase an indicator that may not represent a real problem (an actual security intrusio
90.
▲
by
payne92
9y ago
Finally! Intel has a lot of catching up to do. As GPUs continue to evolve into general purpose vector supercomputers, and as ML/deep learning applications emerge, it seems clear that more and more future chip real estate (and power) w
More ›