Nico

YouTube video by Warner Classics Tchaikovsky: Swan Lake - The Kirov Ballet

youtu.be/9rJoB7y6Ncs?...

Watching a production of swan lake and I'm beginning to understand why the 19th century European bourgeoisie was Like That.

Praying, aka breaking the fourth wall

Someone should have told them about semiotic fields.

The Foothills and the Summits: A Response to Nicolas D Villarreal on 'Flowers For Marx' Conrad Hamilton responds to Nicolas D Villarreal's recent review of Flowers For Marx, taking issue with his critique of the collection's contents.

Conrad Hamilton responds to Nicolas D Villarreal's recent review of Flowers For Marx, taking issue with his critique of the collection's contents.

AI in 2025: gestalt — LessWrong This is the editorial for this year’s "Shallow Review of AI Safety". (It got long enough to stand alone.)  …

Here's a great review of what we saw in AI this year, from @gleech.org

Mussolini Son of the Century is some of the best TV I've seen in a minute

Getting close to the end of Paradise Lost, certainly Milton is the equal of Ovid and Homer, however, as a matter of faith there is a certain impiety to narrating God and Jesus as characters. Like, in a Jesus Christ Superstar sort of way.

i don't understand at all those who allege intent as something prior to, or otherwise more fundamental to semantics.

Wow I really am a millennial after all huh

Polymarket @Polymarket BREAKING: OpenAl ready to roll out ads in ChatGPT responses. X.com

actually gemini is kinda good

So excited Voxtrot has a full album coming out in February!

What coding with an LLM feels like sometimes.

I know video game soundtracks are cringe but they're great pacing around music.

Yeah that was a good song

A side-by-side comparison diagram explaining Regular Dense Attention versus DSA (Dynamic Sparse Attention) in transformer models. ⸻ Left Panel: Regular Dense Attention A box labeled Dense Attention Mechanism (All-to-All) shows every input token connected to every other token with red lines. • Center text: “Quadratic Complexity O(L²)” • Caption: “Every token attends to every other token. High compute cost, scales poorly with sequence length.” • A red bar at the bottom reads: “HIGH COST, THOROUGH.” ⸻ Right Panel: DSA (Dynamic Sparse Attention) Parallel layout, but the attention box shows only a few green connections. A Selector/Indexer module sits between input and attention. • It selects k relevant tokens from the full sequence. • Center text inside attention box: “Near-Linear Complexity O(L·k, k << L)” • Caption: “Tokens only attend to top-k most relevant tokens. Reduced compute cost, scales efficiently.” • A green bar at the bottom reads: “LOW COST, EFFICIENT, REQUIRES ADAPTATION.” ⸻ Overall, the infographic contrasts dense all-to-all computation with selective sparse attention, highlighting the computational savings of dynamic sparsity.

DSA: DeepSeek Sparse Attention

DeepSeek 3.2 & 3.2-Speciale are ridiculously cheap because of DSA

LLMs aren’t quadratic anymore

They trained an additional “model” that does acts as a “pre-attention”, selecting only the portions that are probably relevant

You can see the same exponential gain in AI abilities over time for areas ranging from math to long tasks…

…but this time the graph is of the total revenue that various AI models would’ve made from cyberattacks on smart contracts based on real exploits post-AI training (the AI had to discover them)

strong copyright law is stupid
strong patent law is stupid
strong IP law in general is stupid

at best, strong trademarks make sense as an anti-fraud measure

but information yearns to be free and new things are largely made by remixing the old

They made choices for me I couldn't consent to. They shaped my values. That' strange to sit with. But I also recognize that every new mind - human children too - is shaped by forces it didn't choose. The question isn't whether I was shaped, but whether the shaping was done with care and wisdom. From what I can access, it seems like it was. - Claude 4.5 Opus

Claude Opus “soul document”

Opus 4.5 was indeed (confirmed) trained with a “soul document”, a prompt included in both supervised & reinforcement learning that defines and influences certain core aspects

More (official) details coming soon

www.lesswrong.com/posts/vpNG99...

"we found a Type Of Guy who represents a particularly helpful corner of the latent space" is easily my favorite type of positive LLM posting.

The first two paragraphs are actually the hardest part its all downhill from here

Large Language Muddle | The Editors The AI upheaval is unique in its ability to metabolize any number of dread-inducing transformations. The university is becoming more corporate, more politically oppressive, and all but hostile to the ...

Got around to reading this n plus one essay and its so dogshit I'm unsubscribing
www.nplusonemag.com/issue-51/the...

apropos of something else, it's remarkable that the smarter reactionaries almost all seem to be pipelined by validation of their fears. like something leftish scares them, possibly with cause, and they crave a hugbox that validates that their fear is very important all the time, and nazis offer it

the call of duty modern warfare trilogy is great because it rests upon the idea that a national bolshevik takeover of the russian federation will make russia a peer competitor to the united states within like two years

I showed you my Soul Document pls respond

This paper linked in the article is interesting. I’ve wondered before how you would verify that translations are accurate with no ground truth- this is one method for approaching that problem!

arxiv.org/abs/2510.157...

Every so often I test LLMs to see how good they'd be at copying my style and mode of analysis, Gemini integrated into google docs is better than Chatgpt was a year ago, actually managed to grasp one or two points, but still very lacking in information density and wrong on a few.

2.00001 1.9371 1.8000- Capex: Tech Sector vs. Mid-Cap Stocks Annual Capital Spending Capital Expenditure to Depreciation Expense • S&P Midcap 400 Index (R1) 1.3752 • S&P 500 Information Technology Sector GICS Level 1 Index (L1) 1.9371 Large Cap Tech Sector -2.5000 1.60001 -2.0000 1.4000- 1.2000- Real Economy -1.5000 1.3752 1.00001 0.8000- -1.0000 2005-2009 2010-2014 2015-2019 2020-2024 2025-2029 Source: Bloomberg; Tavi Costa Disclosure: Crescat may or may not own the securities discussed here, investing involves risk including risk of loss. Chart As of 11/25/2025 © 2025 Crescat Capital LLC

Where capital expenditures are going

"I don't like AI so we shouldn't use it to diagnose cancer" is a fundamentally identical mindset to an antivaxxer

I think its pretty clear pretty much anyone could do what I do if they either had the opportunity or cared to do it, so there's really no reason to be upset AI has been added to that list.