Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
rexreed
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
25 ms
·
301.
▲
by
rexreed
4y ago
How much does the typical parked domain make? What is the biggest revenue generator? Newsy seems interesting but is it really working well for your 50 domains? Is it anything more than RSS aggregator with autoposting? What else do you do to
302.
▲
by
rexreed
4y ago
Is it important for podcasts to have guests? Are podcasts that provide insight or discussion or deep dives into ideas without guests not as valuable as ones with guests? Do you primarily seek out podcasts that have guests?
303.
▲
by
rexreed
4y ago
The "FAANGs" or "MAFANGs" of today are the Standard Oil / US Steel / Westinghouse / etc of yesterday. Bezos, Musk, Pichai, and all would fit in really well with Rockefeller, Astor, Carnegie of yesterday. W
304.
▲
by
rexreed
4y ago
A VC once told me that the best thing that Boston Dynamics makes are viral videos. Founded in 1992 with three major corporate owners who couldn't figure out how to monetize their R&D. As a University lab it would be grant funded.
305.
▲
by
rexreed
4y ago
What is your business model? How are you deciding who gets Beta access? What does the voice generation interface look like?
306.
▲
by
rexreed
4y ago
The average person doesn't have room for a sufficient sized chicken coop. The average person doesn't have the wherewithal to get eggs to market. The average person can't efficiently produce eggs in any sort of quantity to sat
307.
▲
by
rexreed
4y ago
Cream prices have more than doubled. At Target, the 32oz Heavy Cream that was $2.99 in Oct 2021 is now $6.49. And there's a cream shortage in any case. Inflation and supply chain issues.
308.
▲
by
rexreed
4y ago
Thanks for the response. I understand that fraud is a very real thing and that crooks will do anything to work the system, but the crooks have very little to lose and the merchants have so much to lose, and Stripe wields a lot of power in b
309.
▲
by
rexreed
4y ago
For sure I read it. In fact, he defines On-Demand Learning as follows in that article: "The second category is "on-demand learning", this is the act of acquiring knowledge when the situation demands it, not sooner nor later.
310.
▲
by
rexreed
4y ago
Says the guy who right on his home page [0] says: "I love making fun small projects and learning new things in the process. I’m interested in computer science, physics, math, and pretty much everything else." [0] https:/
311.
▲
by
rexreed
4y ago
It's one of those ironic Meme shirts sold on Etsy. Same dude sells Lehman Bros Risk Management (2008) shirts.
312.
▲
by
rexreed
4y ago
"I wasn't intending for my arrow sticking out to poke people in the eye. People are just getting taller these days"
313.
▲
by
rexreed
4y ago
This from a fellow who posts on Twitter using his Mastodon Social account handle. I don't get the point.
314.
▲
by
rexreed
4y ago
Exactly. The cloud orientation that differentiates from a non-cloud orientation is not simply a matter of execution abstraction -- that is already handled through libraries and scripts, etc. What would be cloud-oriented are the economic asp
315.
▲
by
rexreed
4y ago
Twitter and Facebook could disappear tomorrow and I wouldn't care at all. Folks should rediscover the personal empowerment of blogging, and the joys of the early days of the 'net when the power was much more widely distributed. Wh
316.
▲
by
rexreed
4y ago
A cloud-oriented language should have a way to output for every function that has financial cost, there should be a cost estimator or inspector that provides estimated cost ranges for execution, or at least some indication of potential cost
317.
▲
by
rexreed
4y ago
"fine-tuned on images of spectrograms paired with text" How many paired training images / text and what was the source of your training data? Just curious to know how much fine tuning was needed to get the results and what th
318.
▲
by
rexreed
4y ago
"fine-tuned on images of spectrograms paired with text" How many paired training images / text and what was the source of your training data? Just curious to know how much fine tuning was needed to get the results and what th
319.
▲
by
rexreed
4y ago
What sort of setup do you need to be able to fine tune Stable Diffusion models? Are there good tutorials out there for fine tuning with cloud or non-cloud GPUs?
320.
▲
by
rexreed
4y ago
I think the intent is to reduce the volume to near 0 during ad and DJ breaks, not normalize the volume against the music volume
321.
▲
by
rexreed
4y ago
Fortunately you can easily override the UA string in Lynx: https://electrictoolbox.com/lynx-user-agent/ lynx -useragent="Mozilla/5.0 (compatible; Googlebot/2.1; + http://www.google.com/bot
322.
▲
by
rexreed
4y ago
If you want fast and minimal, give Lynx a try. The original fast, minimal, privacy protected browser: https://lynx.invisible-island.net/current/
323.
▲
by
rexreed
4y ago
Here's what I learned in life that's applicable to many areas, this being one: Not sucking isn't enough. Even being "Great" isn't enough. And even being "amazing" just isn't good enough to be suc
324.
▲
by
rexreed
4y ago
Do you find any wild animals (raccoons) eat the fish? Also any opinions on floating plants (water hyacinth, water lettuce, frogbit) instead of Taro? I built a pond and put in a soft PVC liner instead of one of those preformed ones. I had it
325.
▲
by
rexreed
4y ago
"They’re hard, but sold as something that everyone should do, so everyone does them poorly, and most people don’t see any value as a result." OKRs are inane management fad. OKRs are "hard" because Objectives are vague an
326.
▲
by
rexreed
4y ago
I remember in another thread you mentioned that you're doing YouTube pretty much full time but you also said ad revenue isn't that much, so I'm still curious if it really is worth it financially and if so, what the best monet
327.
▲
by
rexreed
4y ago
For 2) you mean you loaded it from the adversary's proxy site, just to clarify?
328.
▲
by
rexreed
4y ago
When I saw the word "cavitation", submarines was the first thing I thought about. Blame the Hunt for Red October for that word association. You'd think that this approach to propeller design would be adopted due to quieter ru
329.
▲
by
rexreed
4y ago
Give ParROT13 a try
330.
▲
by
rexreed
4y ago
What exactly are people expecting? All transformer models, of which ChatGPT is just one big fancy example, are just pattern matchers based on a large corpus of text trying to find the next string that completes the pattern. There's no
More ›