BoorishBears

9 ms·

1.
by BoorishBears 7d ago
This seems weirdly pessemistic: frontier labs have much stronger pretraining than most open weights models And reading the release it feels very obvious this is also a ton of aligning their data mix with coding and science: we don't kn
2.
by BoorishBears 7d ago
I don't think HN staff would ever do that, but what was the topic?
3.
by BoorishBears 13d ago
This isn't the gotcha that you think it is: AGI's original definition is being able to do any task that requires human intelligence. The unlock isn't AGI smart enough to invent quantum mechanics, it's suddenly being able
4.
by BoorishBears 13d ago
Cool. Being sole proprietor of AGI 15 years ago should result in monuments and religions devoted to you today. Cancer should be cured, and we should be a post-quantum interstellar fusion-powered civilization. I wish the AGI crowd would fina
5.
by BoorishBears 14d ago
After they buy AZ, making this name foreshadowing
6.
by BoorishBears 14d ago
Would be surprising if it's not on all 3 major clouds soon enough because that's been their general strategy since said break up
7.
by BoorishBears 16d ago
Lol we got literally the opposite: > *Fewer progress updates during long tool runs.* > The model writes less user-facing text between tool calls, especially at higher effort. Set thinking.display to "updates" (beta) to recei
8.
by BoorishBears 16d ago
At least half the changes are just anti-distillation strategies...
9.
by BoorishBears 20d ago
4.7 and 4.8 are smaller than 4.5 and 4.6 (And 4.5 was smaller than 4)
10.
by BoorishBears 23d ago
As the other comment points out, if you've installed more than a few, you've definitly connected to the CAN bus. Few headunits will ask you to connect to the OBD2 port for practical reasons, but the miscilanous manufacturer specif
11.
by BoorishBears 26d ago
I was confused why I was already on the list... this is from 2023.
12.
by BoorishBears 28d ago
I noticed my MBP had a periodic stutter sometimes: every 3 seconds or so the mouse would lock up for a few milliseconds. I tracked it down to American Express's login page (auto loads if you leave the site idle) relying on Akamai'
13.
by BoorishBears 1mo ago
To be really clear, Fable is Opus. Anthropic finished a new pre-training run, Opus-sized models got enough of a jump they could have released Fable as Opus 5... but the economics of Opus models weren't where they wanted. Being the ma
14.
by BoorishBears 1mo ago
If no hand model is what's letting them sell me a high quality mouse for $45 in a two pack... I'll take whatever hands they can get
15.
by BoorishBears 1mo ago
MX4 finally got me looking elsewhere until I stumbled upon this: https://ergodriven.com/products/the-vertical-handshake-mouse I got it specifically because they had a mini rant about the weight of switches that other v
16.
by BoorishBears 1mo ago
OpenAI's moderation API is multi-modal and free with no strings attached in a way that truly boggles the mind. I've put easily over a billion requests (>$100,000 by typical moderation API pricing) through it over the last few y
17.
by BoorishBears 1mo ago
For those curious, peak valuation was Dec 2021: $11.7B
18.
Bending Spoons entered definitive agreement to acquire Airtable for $1.28B(businesswire.com)
29 points by BoorishBears 1mo ago|8 comments
19.
by BoorishBears 2mo ago
You can't beat current API pricing running Kimi K3 a single node, so as I said anyone serious is not doing that for Kimi K3 . Not sure how you're getting "no one serves models on a single node" out of that.
20.
by BoorishBears 2mo ago
Most discourse around GPU prices is nonsense right now. Some people using Spot prices for providers who won't have Spot capacity, some people using hourly rates for instances that are never in stock, some people ignoring commitment dis
21.
by BoorishBears 2mo ago
This is absolutely hilarious https://x.com/CuiMao/status/2049828401201246395
22.
by BoorishBears 2mo ago
I often see this kind of passive-aggressive characterization of the process with AI that's meant to bait people who obviously are expressing their creativity with these models. I hope no one will fall for it: even when you emphatically
23.
by BoorishBears 2mo ago
They do once you're getting margin called.
24.
by BoorishBears 2mo ago
There's something amusingly circular about these conversations, because clearly laypeople like me and the person you responding to are saying "that sounds like it shouldn't be allowed" and the invariable responses are al
25.
by BoorishBears 2mo ago
> Testing whether censorship transmits through unrelated data requires that it never appear in the data. > There was zero China-sensitive content in 220 training prompts, in 176 on-policy training examples, in 181 retained SFT complet
26.
by BoorishBears 2mo ago
> Changing how the model thinks about the Holodomor is completely irrelevant. Your post title is literally "Distilling DeepSeek into GPT-OSS doesn't transfer censorship." Like I'm not really interested in debating you
27.
by BoorishBears 2mo ago
This seems like mildly interesting distillation work wrapped up in a nonsense attempt to drag censorship into the discussion. There's no way your <200 examples for SFT would ever change how the model thinks of Holodomor unless you&#
28.
by BoorishBears 2mo ago
Not surprising it's a hard cutoff: they almost certainly have two infrastructure configurations for the two max sequence lengths Fewer nodes dedicated to prefill per instance, and fewer nodes in total since you don't need to suppo
29.
by BoorishBears 2mo ago
I prefer its sibling library Kotlin, from the makers of the world famous Java IDE: Jetbrains Fleet
30.
by BoorishBears 2mo ago
"the premium of in person" is a string of words that only means something to people in a very specific circle where everyone is constantly trying to tastemake and write in lower case. Ironically they have extremely limited influen

More ›