Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
senko
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
34 ms
·
691.
▲
by
senko
2y ago
Simonw estimates it'd cost less than $10 to categorize 67k+ photos using Amazon Nova: https://simonwillison.net/2024/Dec/4/amazon-nova/#gamoa I agree it'll still be cool to be able to do it all
692.
▲
by
senko
2y ago
Mostly agree, but this is not entirely correct: > Every EU company also has access to the entire continent's consumer base – 450 million residents, which is 50% more than the US. This benefits many EU tech companies – Spotify, Revol
693.
▲
by
senko
2y ago
I think you're both correct. EU Single market is a massive improvement, but it's not yet really single market like in the States (although there can be a bit of legislation/tax barriers between states there as well, depending
694.
▲
by
senko
2y ago
While the underlying mechanics are similar (pension funds), the average European (at least in my corner of the EU) is completely unaware. They know some (government-mandated) amount is withheld and paid out to the government (which then r
695.
▲
by
senko
2y ago
Some live who deserve to die. And many that deserve to live lose their life. Can you give it back to them? If not, don't be so hasty to take it away from others. --Gandalf
696.
▲
by
senko
2y ago
Too late to edit, but here's a great, really in-depth post about using LLMs as judges to evaluate LLM outputs (when you don't have the ground truth for everything): https://cameronrwolfe.substack.com/p/finetun
697.
▲
by
senko
2y ago
When you're writing a test, you usually know the correct answer for that specific combination of input parameters.
698.
▲
by
senko
2y ago
Probably non-optimal long term, but you can easily whip up rudimentary tests with your LLM/lib of choice (see my other comment at https://news.ycombinator.com/item?id=42322794 ) without using any of the eval platforms.
699.
▲
by
senko
2y ago
The output of an LLM is often qualitative, not quantitative, and to test that, you need something that can judge the quality. You're not debating philosophy with the LLM, you're just asking it if the answer matches (semantically
700.
▲
by
senko
2y ago
> The game theory here is that if you're a foundation model "company", you're dead - big tech will kill you. You don't have a product, and you're paying a lot to do research that isn't necessarily tied
701.
▲
by
senko
2y ago
I used it (different lib, same purpose) at a previous startup, shared web whiteboard, to save the whiteboard to pdf/image. We had a rendering engine backed by canvas, so it was easy to just redirect it to this to get the file output. (
702.
▲
by
senko
2y ago
I happen to agree with most of what he said, but man, that tone was insufferable ... (non-wrapping extra long lines have not helped but maybe that's on my user agent) My rules of thumb: * informational website / blog - static fron
703.
▲
by
senko
2y ago
He's absolutely very enthusiastic about many things! I take that into account and mentally adjust when listening to him, but he's been great as [one] source of AI product news.
704.
▲
by
senko
2y ago
I follow these: * Matt Berman on X / YT * AI-summarized AI news digest: https://buttondown.com/ainews by swyx * https://codingwithintelligence.com/about by Rick Lamers Then I manually follow up to lear
705.
▲
by
senko
2y ago
> they're in person 3 days a week. > I asked about remote and was told the company was up to 4 days a week remote. Wait, do they work 7 days a week? If that's their attitude (and I've seen companies where it is), that&#
706.
▲
by
senko
2y ago
I would agree, but also LEGO of today is not like LEGO from 30 years ago. Kids today almost never get raw brick sets and let their creativity loose. Instead, the focus is on specific sets, which you complete, then play around with as ordina
707.
▲
by
senko
2y ago
> if you're going to be using django-ninja (pydantic) This assumes that people don't do multi-page apps or sites any more, which ... isn't true. And I believe django-ninja replaces forms/serialization/deserializa
708.
▲
by
senko
2y ago
Why not? If we have Disneyland, LEGOland, why not MinecraftLand? Minecraft as a brand is now much more than just a blocky PvE game. I have young kids and they'd love to visit a Minecraft themed park. (So would I!) How it'll turn o
709.
▲
by
senko
2y ago
Starlette and Django are wildly different. Django has a batteries-includes approach, benefits from tighter integration of orm, auth, form handling, etc, and has a huge 3rd party ecosystem. First-class async support in Django allows Django u
710.
▲
by
senko
2y ago
"Hello Jane, please have a seat. We need to talk about your productivity. Yes, I know you helped the team through a crunch and delivered the new feature, which works flawlessly and is loved by our users. And our balance sheet is much h
711.
▲
by
senko
2y ago
There are a number of agentic systems that can develop more complex solutions. Just a few off the top of my head: Pythagora, Devin, OpenHands, Fume, Tusk, Replit, Codebuff, Vly. I'm sure I've missed a bunch. Are they good enough t
712.
▲
by
senko
2y ago
No. The scaling laws may be dead. Does this mean the end of LLM advances? Absolutely not. There are many different ways to improve LLM capabilities. Everyone was mostly focused on the scaling laws because that worked extremely well (actuall
713.
▲
by
senko
2y ago
As others commented, the article doesn't really talk about edtech, but about introducing smartphones/tables/computers in curriculum in a way that makes it harder to limit their use both at school and at home. From my kids
714.
▲
How to Give and Take Advice
(senkorasic.com)
1 points
by
senko
2y ago
|
0 comments
715.
▲
by
senko
2y ago
> As Mach was conceived and mostly designed by an academic with no real world industry experience in shipping kernels. You may be thinking of Andrew S. Tanenbaum, who created MINIX, and was famously blasted by Linus for not having indust
716.
▲
by
senko
2y ago
My 2c as someone who worked on a similar product: > it seems like it would be more effective to learn the skills you need rather than using this for a decade. Think of it as a calculator. You do want to be able to do addition, but not ne
717.
▲
by
senko
2y ago
> My worst experiences universally have always been python projects. I don't think I've had a single time where I cloned a python project and had it just work. I'm curios if you can spot a pattern in the platform (win/
718.
▲
by
senko
2y ago
Case in point: I just bought Diablo 3 (released in 2012) on Switch the other day. I'm sure D4 is more modern, but the difference from D3 is nowhere near D2->D3 for the same time span (12 years).
719.
▲
by
senko
2y ago
Hello from Europe. Tried that, didn't work. It's incredibly hard to quantify "low-performing" for white-collar workers, because any measure is either easily gamed, actually creates roadblocks and bad incentives, or both.
720.
▲
Jupyter Notebook Setup for Research and Production
(goranperetin.com)
1 points
by
senko
2y ago
|
0 comments
More ›