Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
dmos62
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
271.
▲
by
dmos62
9mo ago
Warsaw is top 15, Krakow and Warsaw are the only European cities in the top 15. For some added context, it's around -10 degrees celcius there right now. I don't know why Poland stands out here, but I know that older residential ar
272.
▲
by
dmos62
9mo ago
Will try this out. Was always envious of how Augment was able to do this. Kudos.
273.
▲
by
dmos62
9mo ago
Have been having fun on here for some 10 minutes now. Not interested in the gamification, but I like the flash-card-esque, short lessons and the interactive tests.
274.
▲
by
dmos62
9mo ago
Retrieval-augmented generation. What you described is a perfect example of a RAG. An embedding-based search might be more common, but that's a detail.
275.
▲
by
dmos62
9mo ago
Oh, I didn't get that! That's pretty witty indeed. Now I feel bad.
276.
▲
by
dmos62
9mo ago
What use-cases gave you disappointing results? Did you build some kind of RAG?
277.
▲
by
dmos62
9mo ago
Ah, yes. Post is an LLM-something project: top comment is a general critique of LLMs. Waiting for this to get old. Meanwhile, at least you get points for being funny.
278.
▲
by
dmos62
9mo ago
That's more like it.
279.
▲
by
dmos62
9mo ago
You complain that it doesn't explain this or that, but you can just ask it: it's a discussion or a dialogue: what books can say that? I'm not discounting good books or good educators, it's just that LLMs are good too, I
280.
▲
by
dmos62
9mo ago
I think it's good to explicitly point out that low-effort participation is discouraged on HN. It might not affect the actual "offender", but it's meaningful to the rest of the readers.
281.
▲
by
dmos62
9mo ago
Feel free to make one. Would you expect someone to cry "propaganda" if you did?
282.
▲
by
dmos62
9mo ago
That's true, and it's worth mentioning that both your parent and grandparent comments are painfully low-effort.
283.
▲
by
dmos62
10mo ago
Those are great links.
284.
▲
by
dmos62
10mo ago
> zero value to me in learning the subject Perplexing how different our perspectives are. I find this super useful for learning, especially since I can continue chatting about any and all of it.
285.
▲
by
dmos62
10mo ago
>The fact that Ÿnsect failed doesn’t mean the entire insect farming sector is doomed. Competitor Innovafeed is reportedly holding up better, in part because it started with a smaller production site and is ramping up incrementally. >F
286.
▲
by
dmos62
10mo ago
This discussion might be a bit more grounded if we were to discuss a concrete LLM response. Seems pretty freaking good to me: https://chatgpt.com/share/6955a171-e7a4-8012-bd78-9848087058...
287.
▲
by
dmos62
10mo ago
I consider the "perfect fortgetfulness" of LLMs a great feature, because I can then precisely select what the context is for a given task. Context is additive, so once something's in it, it's doing something: most I coul
288.
▲
by
dmos62
10mo ago
Anyone's aware of something similar for making interactive (or video) tours of apartments from photos?
289.
▲
by
dmos62
10mo ago
Is Facebook US-centered? I don't think that means much.
290.
▲
by
dmos62
10mo ago
In my case, the business did have human support assistants, but didn't do reservations via the phone. I had to switch to the web app for that, which was annoying (I was driving?). I guess doing user identification over the phone and sc
291.
▲
by
dmos62
10mo ago
That's not true. I recently called to make an appointment. I don't care if it's an AI. I would actually prefer it, because I wouldn't feel bad about taking a long time to pick the best time. Don't you think you'
292.
▲
by
dmos62
10mo ago
Free software is in everyone's interest. It's just that if you drag your feet, probably someone else will pay for it.
293.
▲
by
dmos62
10mo ago
Good customer support lines? Is there a reason why it can't provide good support. I often use chatgpt's voice function.
294.
▲
by
dmos62
10mo ago
> Agent Framework offers two primary categories of capabilities: > AI agents: Individual agents that use LLMs to process user inputs, call tools and MCP servers to perform actions, and generate responses. Agents support model provi
295.
▲
by
dmos62
10mo ago
His assistant may have used his account to place the order.
296.
▲
by
dmos62
10mo ago
I'm sure our viewpoints are more similar than it seems, and we eventually would find a fairly spacious middle-ground, but I'd prefer not to continue: thanks for the conversation.
297.
▲
by
dmos62
10mo ago
> I was specifically talking about accidental shootdowns of civilian airliners. Maybe you should think twice before dropping offhand comments related to mass killings. If you want to talk about what exactly I said that's "polit
298.
▲
by
dmos62
10mo ago
Had to look up what vm gaming is. What's your motivation? If you don't mind sharing.
299.
▲
by
dmos62
10mo ago
You think I think there isn't a difference between bias and prejudice, while I think you think there isn't a difference between prejudice and knowledge. What I really care about is guilty-until-proven-innocent masquerading as civi
300.
▲
by
dmos62
10mo ago
Guy, you're the only one here acting like this is a competition. Do you think what Russia does is somehow more acceptable if you can find other criminals? Yeah, just lean into it. Good luck with that.
More ›