k__

3 ms·

1.
by k__ 5d ago
Half-OT: Is Jev a decoder (e.g., BERT) or is it some kind of encoder (e.g., GPT) that just happens to be trimmed down to only outputting a handful of tokens for the answers and their probability?
2.
by k__ 6d ago
Fair, but Y isn't used much in German.
3.
by k__ 6d ago
That's how they're sold, isn't it?
4.
by k__ 8d ago
I think so, yes. However, it might have fewer restrictions than a BERT and/or is smarter (whatever that means).
5.
by k__ 9d ago
Does that mean, it's not compressed anymore?
6.
by k__ 11d ago
I didn't have the impression anyone felt ashamed of dating online back in the days and I don't have the impression vibe coders feel any shame today.
7.
by k__ 12d ago
I was yesterday years old when I learned that those open weight models need custom code to run. Somehow I expected inference engines are generic LLM runtimes that can execute any weight. So, to get this right. Someone trains a model. They r
8.
by k__ 16d ago
Update: I'm using it right now and it's noticeably faster. I'd also say, it seems smarter, but I think that's because of some harness updates I installed. (I haven't used pi for almost a month)
9.
by k__ 17d ago
So, while the throughput was 400-500tps in beta its now ~150tps on OpenRouter. I was hoping for a bit more, but it's still 100% faster for a very good price, so I won't complain.
10.
by k__ 17d ago
Nice form factor. I have a Samsung Fold 4. The front screen is too narrow and the folding screen doesn't open all the way anymore. Not what I expected from a 1400€ phone.
11.
by k__ 17d ago
At least they can use another provider or self-host.
12.
by k__ 17d ago
Beta testers report >400 TPS. https://www.geeky-gadgets.com/deepseek-v4-1-flash-review/ I hope some of those speed increases will make it to production.
13.
by k__ 17d ago
I used flash with pi and it worked pretty well. It built this whole IaC plugin from scratch: https://github.com/fllstck/nebius-alchemy
14.
by k__ 18d ago
I don't know about animation, but it seems the UI system and their widget library was overhauled. "The Feathers widgets have migrated to BSN, Bevy's next-generation scene system. BSN is a better foundation for widgets than t
15.
by k__ 18d ago
Yeah, I think only in crypto the non-us providers have a leg up, because running a node just needs an instance.
16.
by k__ 20d ago
Yeah, there are plenty of slow burns. Yet, I wouldn't go so far as to say every novel is one. If you aren't an established author, publishers won't take on your manuscript if it doesn't hook them with the first chapter,
17.
by k__ 20d ago
Interesting that he has this view. I would have taken him for mostly pro LLM.
18.
by k__ 23d ago
Any signal that allows the model to see what's wrong helps. Checking code is (relatively) easy, you can use static type checks, linters, and execute it to see if it's correct. Fact checking is harder. A RAG can only check what&#x2
19.
by k__ 24d ago
I had the impression that GLM 5.3 is just open weight and not open source.
20.
by k__ 24d ago
I use a Google sheet.
21.
by k__ 24d ago
Somehow RRS removed their search function recently. Whelp, as with all those aggregators, it was nice while it lasted.
22.
by k__ 29d ago
I can't even login to languagereactor.
23.
by k__ 29d ago
And with exponential, you mean sigmoid.
24.
by k__ 29d ago
Cool idea! While I'm not so sure about learning a language with content from fictional sources, that should at least make it more relevant for the learners. And you can use it on podcasts too, where people talk normally. The biggest hu
25.
by k__ 1mo ago
The "marked" consolidated.
26.
by k__ 1mo ago
I'm always amazed by solutions for user interactions. I once built a Bomberman clone that had control issues. The first iteration of the controls were clunky, because I only checked if a player was touching a block in the direction the
27.
by k__ 1mo ago
I'm working remote since 2014. I built my whole life on that premise, traveling through the country on a weekly basis. I love the freedom.
28.
by k__ 1mo ago
Just a few days ago, I received one because of "the odour nuisance is no danger for the population" People were calling the emergency hotline because the wildfires 200km away got so big that their smoke engulfed the city.
29.
by k__ 1mo ago
I had the opposite impression. LLMs can abstract pretty well. They write quality code for obscure frameworks no problem, as longs as you take the time to set up tight feedback loops.
30.
by k__ 1mo ago
Yeah, None of them offers the same cache hit prices and my cache hits are >95%.

More ›