refulgentis

5 ms·

1.
by refulgentis 5d ago
I don't get it, why?
2.
by refulgentis 7d ago
I'm sorry, it's going over my head still - my reading is "all models with any training after August 10 know how GPT 5.5 Pro thinks", but I'm not sure why - my initial guess was that's when GPT 5.5 was released,
3.
by refulgentis 7d ago
The thoughts trick was known before their paper / August. I "independently" "invented" it for the first Anthropic reasoning models because the API required you have thoughts for each assistant message. My app lets y
4.
by refulgentis 8d ago
I'm less bullish, this model and previous are halfway decent and you can do diffusion sub-agents and sub-agent-agents today, and there hasn't been a sea change, or anything noticeable or well-known. Economically, there's ~no
5.
by refulgentis 8d ago
Carefully worded; it's extremely likely to be the same large frontier model that started training again on August 28th as well, as they revealed in some of the RL message board follow-up - for several reasons, most importantly, if we a
6.
by refulgentis 8d ago
https://news.ycombinator.com/newsguidelines.html - "Be kind. Don't be snarky. Converse curiously; don't cross-examine."; "Don't be curmudgeonly." FWIW, with a foothold in both domains, thi
7.
by refulgentis 12d ago
Is any piece you've seen in good enough shape to be in a Lean library?
8.
by refulgentis 13d ago
If it makes you feel any better the curmudgeon who drives the ARC-AGI tests feels the same way, that's why we're on 3 and I'm sure we'll see 4. Also, we can all avoid calling it "moving the goalposts" so no one
9.
by refulgentis 14d ago
I don’t know why people think DeepSeek did reasoning models / RLVR before OpenAI, there was a gap of months.
10.
by refulgentis 14d ago
There's, quietly, a llama.cpp WebGPU backend that works *great*. Some hacking required, it's unsupported, a side project for one of the lead maintainers and someone in school. Note of caution, llama.cpp isn't what it was, the
11.
by refulgentis 14d ago
They're quite selective in benchmarks, c.f. notably only bad one is 10% on TerminalBench. It's a really addled model, one time I said "Hi" and it built out a 4 panel hello world app with (fake) weather, a todo list, and
12.
by refulgentis 14d ago
Project is de facto dead, used it for many years and had to rip it out 6 months ago, don't waste your time.
13.
by refulgentis 15d ago
Let’s say it’s fine to give emotionally based arguments the same credence as rational arguments. That’s a huge cede, and yet, it doesn’t change anything, let’s see here: If my cousin kept ranting about my other cousin was going to go bank
14.
by refulgentis 16d ago
My $0.02: it doesn't actually exist. GitHub repo indicates they're working on basic, basic, bringup. Like, Objective-C works at all and CLI utilities work . This is gutting, when combined with a prior post showing this dates ba
15.
by refulgentis 19d ago
Running on ported userspace components, with a shim, isn't the same as running on the OS in itself. That really matters, some times. But not the vast majority of the time. (source: been coding iOS since pre-SDK iPhone OS 1.0)
16.
by refulgentis 19d ago
There’s this strain of thinking that’s complete alien to me, I can’t interpret what’s being said and it sounds incredibly aggressive. ex. our lead post asking if Sam Altman still thinks GPT-3 is unsafe to release, and I don’t understand wha
17.
by refulgentis 19d ago
> Not even sure why this person is trying to get this stuff to run on hardware that wasn't designed for AI? Posturing/overclaiming like this shades rather than illuminates, there are no worlds in which the "M3...wasn'
18.
by refulgentis 19d ago
Slop; all those LocalLlama threads "you" read had #s too. And Q1 quant? Why? You have the memory...
19.
by refulgentis 21d ago
Dario has less space than a Nomad!
20.
by refulgentis 22d ago
Huh, makes 0 sense to me, blood tests w/o much blood isn’t same as glucose checking, we’ve been able to do that w/o much blood for…a long long time.
21.
by refulgentis 22d ago
Who?
22.
by refulgentis 23d ago
The person you're replying to is saying "Sites I use block all VPNs besides Apple's subscription service VPN" - I'm not sure they're not blocking it just because they fervently believe in Apple's privacy c
23.
by refulgentis 23d ago
A guid isn’t a user id, if it is in this case, it’s an abuse of how guid is used, at least colloquially. I haven’t read enough to understand if it is user / machine id, I.e. only globally unique in the sense it’s a globally unique en
24.
by refulgentis 23d ago
You are misunderstanding, and, they are not downplaying. A revision or other reply would benefit from explaining how their value chain x location correspondence is wrong - perhaps missing a step? Though, in anticipation of a revision, I won
25.
by refulgentis 25d ago
c.f. https://news.ycombinator.com/item?id=49403836 , then from there, you'd need to see a couple of orders of magnitude before it's tractable for LLMs. Bottom of page linked from HN (currently https://ww
26.
by refulgentis 25d ago
My 15 second read of just the front page aligned with you, but when I saw replies pushing back, I went back and loaded News, then cross-verified some of the claims. It's real.
27.
by refulgentis 26d ago
On their API? Almost definitely, China is good GPU constrained. On the technical aspects? Absolutely not, GLM 5.3 is a relatively small MoE.
28.
by refulgentis 29d ago
Bloviating nonsense, 3rd time I’ve seen something like this in HN since TurboQuant came out. You don’t need float32, never did. Source: I’ve been writing on device embedding code for 4 years.
29.
by refulgentis 1mo ago
It’s right there!
30.
by refulgentis 1mo ago
I flagged this, so I can act as an interlocutor for the shadowy n’er do wells flagging you, if you’re interested.

More ›