Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
joegibbs
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
31.
▲
by
joegibbs
2mo ago
The concept of “doxing” has come too far. You used to have a phone book that had everyone’s names, addresses and phone numbers in them. Now it’s treated as if there’s an expectation that someone’s address is very private information.
32.
▲
by
joegibbs
2mo ago
What are you missing from a Range Rover if you get one of these? They’re probably more reliable. They just need to work on the terrible names - Geely, Zeekr, Jaecoo? They sound like they smashed a keyboard and then maybe added some vowels.
33.
▲
by
joegibbs
2mo ago
The hard part is that it locks you in to whatever people believe at the moment, which might be proven wrong in the future
34.
▲
by
joegibbs
2mo ago
It’s going to do nothing. Remember that old screenshot of a Steam group “Boycott Modern Warfare 2 - we want dedicated servers!” Where almost everyone is currently playing MW2?
35.
▲
by
joegibbs
2mo ago
I think this whole piece is just a collection of lazy cliches. - Rug pull scam What's going to happen when the AI companies jack up prices 20x to meet API prices? 1: they won't, because they'll get their lunch eaten immediate
36.
▲
by
joegibbs
2mo ago
Rebuttal to what? I’m complaining that it’s a totally pointless comment that could be used on any AI related post. The comment just says “AI is slop garbage that sucks”. What am I going to say to that? Obviously it can’t suck that badly or
37.
▲
by
joegibbs
2mo ago
I might build a Chrome extension that looks for “AI” in HN submission titles and shows a fake comment at the start saying “AI is all slop garbage that’s total slop and I hate it. AI sucks and it’s slop!” It will save people time reading the
38.
▲
by
joegibbs
2mo ago
Momentum from Claude Code's success at the start of the year and hype from Fable I think. It seems like most people I know have switched back to GPT-5.6 (which I find gives much results than Fable, you get more usage), but I imagine th
39.
▲
by
joegibbs
2mo ago
“Claude avoids saying "genuinely", "honestly", or "straightforward"” That’s interesting- they tried training the model to be genuine and honest (I’m not sure it actually made it honest), but now it says “genuin
40.
▲
by
joegibbs
2mo ago
“The [thing that can’t remember] remembers” is a big one. Loves talking about memories and remembering.
41.
▲
by
joegibbs
2mo ago
There’s probably already a LibTouchupGl available since 2005, a perfect touchup tool in 83kb, but nobody had heard of it. On the other hand people can just say “ChatGPT do everything” and it can
42.
▲
by
joegibbs
2mo ago
There are a ton more games now, you could easily get one with similar graphics/gameplay to a 2006 game for much cheaper.
43.
▲
by
joegibbs
2mo ago
“The noun verbs.” is such an incredibly Claudeish sentence. The other models don’t do that stuff. I’d say it’s actually biggest tell for Claude writing.
44.
▲
by
joegibbs
2mo ago
More people are probably using stuff like Claude Code, Codex etc from the CLI or desktop apps, meaning they’re actually using more AI and skipping Zed altogether
45.
▲
by
joegibbs
3mo ago
They were trained on a lot of garbage human-written slop originally, they have to post-train the models so the responses are good. Luckily if you’re an AI company you probably have the whole chat, so it’s easy to categorise it: the user mig
46.
▲
by
joegibbs
3mo ago
With hundreds of millions of people, what are the odds that more die from scurvy than the outbreak?
47.
▲
by
joegibbs
3mo ago
It does, but I don’t think the Ukrainians will get much flak for it. Bombing consumer goods distributors doesn’t worry people as much as bombing food distributors or chemical warfare on farms (eg Agent Orange), and besides that direct strat
48.
▲
by
joegibbs
3mo ago
I don’t think the slippery slope is a fallacy. If a group makes a change in one direction is it not more likely that they will continue making changes in that direction than if they had not made the initial change?
49.
▲
by
joegibbs
3mo ago
GPT-Image has that horrible deep-fried look to it. Can't draw anything without looking like it's got grit thrown over the whole image. Nano Banana is much, much better here, it's often hard to tell if it's fake. Apart fr
50.
▲
by
joegibbs
3mo ago
That legalistic "I can do whatever I want as long as it's not breaking the law" argumnet lasts until the first time they ban wrongfully ban someone for it and get huge negative PR
51.
▲
by
joegibbs
3mo ago
I like the moving cars. They're copies of the stopped ones right? Adds a good effect.
52.
▲
by
joegibbs
3mo ago
I think that's literally about 3-4 requests. Fable is ridiculous expensive on API pricing.
53.
▲
by
joegibbs
3mo ago
Doesn't it say something if 5% of a group are wacko lunatics if, for instance, a random sample of the rest of the population has only 0.5% wackos?
54.
▲
by
joegibbs
3mo ago
The problem is the whole idea of working from first principles to try to discover the actual truth, rather than just going along with other ideas that other people have created through history. The benefit of going along with other peoples&
55.
▲
by
joegibbs
3mo ago
What about an armadillo playing a piano? There are so many potential combinations It would say something if the pelican looked great but the armadillo looked terrible
56.
▲
by
joegibbs
3mo ago
What does it say if you ask it in Chinese? I imagine English training data would be more critical of China than Chinese
57.
▲
by
joegibbs
3mo ago
But who are you to decide whether suburban Americans are living the right kind of life? Surely if they wanted to they could move to an apartment in a dense inner-city area. It would be cheaper than living in a large house in the suburbs an
58.
▲
by
joegibbs
3mo ago
A timer tool with a callback feature would take like a couple of hours to implement, a model which has a native internal ability to know the time would take ages to make
59.
▲
by
joegibbs
3mo ago
Why would the rows be gigabyte size? You’d just save the text.
60.
▲
by
joegibbs
3mo ago
But there must be a ton of generic questions that people ask. Stuff like "What's the capital of country X?" - it's probably at least 10% of queries. Memories, custom instructions etc would invalidate them, but if you can
More ›