Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
ggeorgovassilis
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
9 ms
·
91.
▲
Running an Offline Copy of Wikipedia
(blog.georgovassilis.com)
3 points
by
ggeorgovassilis
3y ago
|
0 comments
92.
▲
Running Wikipedia Offline
(blog.georgovassilis.com)
1 points
by
ggeorgovassilis
3y ago
|
0 comments
93.
▲
by
ggeorgovassilis
3y ago
> I'm still not convinced that this is what FTL travel would look like to an external observer. I'm not either, I'm saying that if that's the case, there is an easy way to find out. > I don't think you've
94.
▲
by
ggeorgovassilis
3y ago
> A skilled data analyst would never even attempt that before having a very strong theoretical model I know for a fact this generalisation to be false
95.
▲
by
ggeorgovassilis
3y ago
> It's also unlikely that we could even measure it for the scenario outlined I understand the LIGO data is available. A skilled data analyst could look at it and tell if it's possible or not. If aliens FTL around our neighbourh
96.
▲
by
ggeorgovassilis
3y ago
> isn’t that relative to your origin It applies to the universal scale, I'm not sure how more absolute it gets. Every observer regardless of frame observes that the expansion is faster than c. > And wouldn’t light then also be tr
97.
▲
by
ggeorgovassilis
3y ago
Some of your observations are good, for those I thank you, and revised the post accordingly. > I wish the article justified this statement Your wish has been granted. Also, it's in the referenced Wikipedia article: "galaxies th
98.
▲
Detecting FTL with LIGO
(blog.georgovassilis.com)
35 points
by
ggeorgovassilis
3y ago
|
22 comments
99.
▲
Ask HN: What is the most complex topic you understand?
4 points
by
ggeorgovassilis
3y ago
|
3 comments
100.
▲
by
ggeorgovassilis
3y ago
I remember that, too. Got such a floppy from a computer magazine. I think one was for the PC and then other for Amiga or Atari.
101.
▲
by
ggeorgovassilis
3y ago
If the topic enthuses me, >12 hours. If it doesn't, 2 + long break + 2 is a max. Advice 1: validate your assumptions. You seem to imply that maximising work hours maximises productivity. It could well be that you are above-average p
102.
▲
Team topologies fix the Spotify model
(blog.georgovassilis.com)
2 points
by
ggeorgovassilis
3y ago
|
0 comments
103.
▲
Transform web content with haproxy and Lua
(blog.georgovassilis.com)
2 points
by
ggeorgovassilis
3y ago
|
1 comments
104.
▲
by
ggeorgovassilis
3y ago
The post describes a (complex) use case of transforming web content with haproxy LUA scripts
105.
▲
Ask HN: Use Surplus Solar Power?
1 points
by
ggeorgovassilis
3y ago
|
0 comments
106.
▲
LLMs and the Extended Mind Thesis
(blog.georgovassilis.com)
2 points
by
ggeorgovassilis
3y ago
|
1 comments
107.
▲
by
ggeorgovassilis
3y ago
TL;DR: LLMs might be able to escape the sandbox
108.
▲
Fictional Cross-Examination of B. Clinton by Alan Shore and Saul Goodman
(chat.openai.com)
2 points
by
ggeorgovassilis
3y ago
|
0 comments
109.
▲
by
ggeorgovassilis
3y ago
> ... just speculation. The formulation (3rd conditional) makes it clear the author is aware they're speculating. > This [the dump is worthless] is entirely incorrect Which isn't telling us why (the dump is valuable). You sa
110.
▲
by
ggeorgovassilis
3y ago
The for-pay package includes web integration, that should work OOTB. Today it insisted the feature doesn't exist. The dialogue demonstrates a feature block override.
111.
▲
Tell HN: I coerced ChatGPT into doing something it thought it couldn't
(chat.openai.com)
3 points
by
ggeorgovassilis
3y ago
|
3 comments
112.
▲
by
ggeorgovassilis
3y ago
TLDR: asked it for a web page summary, it claimed its browsing tool couldn't do it. I then convinced it to try harder. Edit: chatGPT4
113.
▲
by
ggeorgovassilis
3y ago
I'm a long-time, hard-core FF user on all platforms and I admit to having used weird extensions. As time goes by, however, I wonder whether it's worth it. Too many websites break because they aren't tested with FF. I wonder w
114.
▲
by
ggeorgovassilis
3y ago
You cover more points than I had in mind. > If you have experience of shipping desktop apps to end users I used to work on desktop applications in the early 2000s built with VB, VC++ and Java, recently (for fun) with PWAs. > what are
115.
▲
by
ggeorgovassilis
3y ago
> but why do people make businesses to sell them Building something and maintaining it are two different things which require different skills, different finances and different partners. Innovation requires a curious, scientific mind. Ma
116.
▲
by
ggeorgovassilis
3y ago
I understand Youtube uses a variety of codecs and encoding parameters to serve videos to a multitude of platforms and clients. Maybe there are restrictions on some clients that force Youtube to pick a codec for them which makes ad embedding
117.
▲
by
ggeorgovassilis
3y ago
You are probably right, eg Losslesscut in Linux does exactly that: no transcoding, no overlays nor any editing other than cutting and moving around sequences from the same video. Editing and exporting such videos is blazingly fast. Adding a
118.
▲
by
ggeorgovassilis
3y ago
Since ads are personalised, this would produce a new stream per view which would come with a computational overhead. The stream would have to be rendered on the spot and not in advance, which puts further pressure on compute resources durin
119.
▲
by
ggeorgovassilis
3y ago
1) I love the aesthetics and the game play. First time I learned about this was reading a copy of 101 basic computer games during summer break. I didn't have a computer back then but was fascinated by the book, so I pried my parents un
120.
▲
by
ggeorgovassilis
3y ago
I feel stupid and embarrassed after spending a weekend implementing my own elastic-based solution.
More ›