Lockal

8 ms·

1.
by Lockal 4d ago
Related: Ask HN: Can we please limit the AI news flood? https://news.ycombinator.com/item?id=49657850
2.
by Lockal 6d ago
Soon you will see how a rogue state develops a biological weapons using fine-tuned local LLMs (they are already doing it, btw), and all so called "developed" countries can't even research it, because every search engine block
3.
by Lockal 7d ago
> I love that it has a feature to prove images came from the real world Congrats to Apple achieving nation-wide tracking, embedding some kind of yellow dots[1], but this time this time this is based on strong cryptography and non-removab
4.
by Lockal 12d ago
I am so sick of the PR from AI companies. Every single fucking time, they post the same bullshit about the dangers of their own products, just to stay in the media spotlight, steal market share from their competitors, gatekeep competitors o
5.
by Lockal 14d ago
s/text/prose/ s/'/’/ The honest seam is obvious: claudish works; imitation does not. That’s not nothing.
6.
by Lockal 18d ago
The title is straightforward lie. Changing +50% to +25% gives -17% in result.
7.
by Lockal 21d ago
FTFY: https://xcancel.com/hackernews/status/2092359322433191997
8.
by Lockal 22d ago
Frontier LLMs are much better than I am. If you are not asking them they turn into a pumpkin. Babysit them like junior devs (but braindead), pay monthly subscription to model providers, pay for extra tokens, and/or buy RAM$$$ to get 2-
9.
by Lockal 23d ago
Wait until they find that "Russian backdoor" sits inside their government and issues orders to purchase such cameras.
10.
by Lockal 1mo ago
A grossly misleading article; > almost everyone has misaddressed Almost no-one, unless we are talking about LLMs and Google Translate users who could attempt map translations 1-to-1. Oh well, a new AI generated fake piece was published,
11.
by Lockal 1mo ago
As if Anthropic employees use massively different version of model. Not a single load-bearing, epistemic, heavy-lifting, push back, frankly, pivotal, crucial - for 48 pages.
12.
by Lockal 1mo ago
I'm not sure how to properly measure DRAM access, but for register/immediate-only instructions there is https://uops.info/table.html . It does not measure x87 transcendentals (I guess because microcode results in d
13.
by Lockal 1mo ago
"Generate an article about Chinese AI attacking our lab"
14.
by Lockal 2mo ago
This is called Sci-Hub and LibGen
15.
by Lockal 2mo ago
The page is AI slop, don't waste your time
16.
by Lockal 2mo ago
It is like Microsoft publishes a manifesto "Our position on open-source operating systems" in 2026: - Over the /last few days/ there has been a lot of discussion about open-source systems, especially those that are not m
17.
by Lockal 2mo ago
More like "The Emperor's New Clothes" moment, but the Emperor had no clothes from the start. The corruption of 'Open' AI was visible since the start, even before their infamous "new model GPT-2 is too dangerous
18.
by Lockal 2mo ago
AI slop
19.
by Lockal 2mo ago
That is not dumb - erasing copyrights is just their business, and even when caught has zero consequences (for them). Providers can use any users input for improving their models, either by consent, or by flagging any dialogue for safety rev
20.
by Lockal 2mo ago
I fed it to Google AI Studio, enabling tool execution and disabling web access. It also quickly verified it with SymPy, then went into psychosis. 5 minutes later: all previous chats are loading fine, but the only "Counterexample to the
21.
by Lockal 2mo ago
https://xcancel.com/theo/status/2078217355780624864
22.
by Lockal 2mo ago
More likely that Constitution was simply generated by LLM. Nobody in a sane mind will write 80 screens of test to deliver the idea that nobody asked for (a.k.a. slop). Quick check: Ctrl+F genuine - 50 times, honest - 57 times, ethic[al] - 1
23.
by Lockal 2mo ago
Exactly! This it how isolated hairline strokes look for a bigger image: https://imgur.com/a/UqBdRKU Just a single frame (any frame). After that any modern LLM can read it (if not - single dilate step helps). Funny enou
24.
by Lockal 2mo ago
It is not really "vintage" for modern Russia. For example, search Combined Heat and Power control room images: https://www.google.com/search?udm=2&q=%D0%BA%D0%BE%D0%BC%D0%... - barely any computer screen. I&#x
25.
by Lockal 3mo ago
I'm a simple man, I see no benchmarks at https://arena.ai/leaderboard - I can 100% tell this is a scam.
26.
by Lockal 3mo ago
Should have asked to count R's... In a "strrawberrry".
27.
by Lockal 3mo ago
> Google suspended Fiu’s gmail. Thousands of inbound emails plus rapid API calls triggered their fraud detection That's a good enough reason for me to never run agent on anything else other than burner account. And only if the pla
28.
by Lockal 3mo ago
"We are evolving" is yet another way to say "we are doing mass layoff"?
29.
by Lockal 3mo ago
New "delve" alternatives: honest (4x), real(ity) (11x), genuine (3x). While older LLMs were tuned to slop, current ones are tuned to actively deceive.
30.
by Lockal 3mo ago
Chrome on Linux suffers from this too. The issues are standard for all naive immediate GUI libs: - constant repainting (while literally nothing changes) -> constant CPU load - some libs can skip repaints but have no concept of damage are

More ›