Home Creators Posts Import Register Favorites Logout

Videos

  • 558033045.mp4

Downloads

Content

Recorded on: October 29th,2025 Nvidia Hit 5 Trillion Sam Altman AI Algorithms Argentina Amazon Layoffs

Files

Comments

Pavate

34+33

Gamer Gavin

Aiden gave me the confidence I needed to only cry a little bit when I use the bathroom

Somedoood

I worked at Nvidia - Per Atrioc

John Doe

Notice how much NVidia has grown since Big A left… seems the glarketer was holding them back

Dan

Do people use twitter algorithmically at all? I just see the people i follow and their retweets. Youtube for example I let the algorithm take me for a ride but twitter just doesn't need it imo, Grok stuff aside

Somedoood

Shucks, they didn't talk about how much glizzies 5tn can buy :(. And the other but I guess they chose to ignore. Makes sense why they would do that but, still wanted a response.

Peter Sadek

Very disappointing to see Doug propaganda spread by Doug here. On a serious note. Every social media platform including twitter and YouTube have followed tab next to the for you tab. We already have the tools to fine tune our front page. People just don’t use it.

Ian

Completely agree with Doug here @36:40 I don't trust Elon or Twitter to do this well but this is a concept that may have legs. YouTube is probably the best test case at large scale due to video adding data complexity. And if done well would be a major win for YouTube who has only seen increasing complaints about their algo. Betting Instagram or Threads rips it off first though and it's a mess.

Bailey Yin

Agreed. It has the same feeling as not trusting the current admin to handle some of the actually decent ideas

WillyThaJawn

I think the best outcome from the AI algorithm stuff is that users realize people don't actually talk/act/think the same way they do in person as they do online and that we get away from screens (except to watch LS) and connect with their surrounding communities and environment

Matthew Sand

As someone in the tech field, I think the easiest way to understand decisions being made right now and whether or not something makes sense, is to stop calling it AI. If you just replace whatever you’re talking about with “LLM”, functionally what is being spoken about when speaking about “AI”, a lot more becomes clearer. LLM’s are a way for humans to communicate with computers with natural language rather than UI. That’s the brunt of what they are useful for, and some cool quirks come out of that like being good at writing some “good” code (in practice in industry it’s not much more than better google search) To be clear, they are phenomenal at natural language tasks, but beyond that it’s pretty clear they are going all in on one piece of “ai” and assuming it’s all we need. For example, the twitter algorithm using an LLM to be fine tuned is absurd, not just from a compute standpoint but it’s also just not what they’re good at. You’ll speak to the LLM in natural language and it’ll use the same mechanisms that already exist to change what you see - due to speed and context limitations, LLMs cannot make decisions like what to show you in real time. If all it’s doing is using existing mechanisms, but instead of UI, you speak in natural language, why do these systems not already work / aren’t already implemented?

Billy Hatfield

So funny thing about hiding instructions places, I've seen comments from teacher friends that hide instructions in white text so the students don't see it, but it will mess up the AI when it's trying to generate stuff. It's pretty funny.

Ensyllis

I work with quant models. My personal experience is that it's very hard to find any sources of revenue from our models that don't come from riding the Bubble or insider trading via the current admin. As we have the admin throwing a wrench into the market and more investors pulling their money out, the only assets rising in price reliably are the mag 7, while the companies dropping are dropping due to Trump randomly tweeting on a Tuesday Morning. While we're under pressure to deliver more returns, even our models which are 'correct' isn't delivering returns since other investors don't care about if we predicted a company to earn better their next earnings period. Their attention is on the AI Hype as well as Trump's tweets. Lots of smaller/medium size equity pods are succeeding during this time, but their returns are from profiting off volatility in Market Making. (Buying an Asset at $3.01 selling at $3.03 in the same second) or they're succeeding due to riding the Bubble. We're under pressure to continue deliver returns which forces me to build models that mimic dumber ways people choose to invest. If you're interested: https://www.businessinsider.com/quant-hedge-funds-losses-summer-cubist-qube-man-group-2025-7 https://www.businessinsider.com/hedge-funds-flagship-vs-smaller-funds-multistrategy-2025-10 https://www.bloomberg.com/opinion/articles/2025-08-04/hedge-funds-are-hurting-from-a-garbage-rally-who-s-to-blame And if you want reddit insights: https://www.reddit.com/r/quant/comments/1m8dq8z/quant_hedge_funds_suffering_mystifying_string_of/

Kevin Jamer

You guys need to get one of the crypto sharks on. The mooch would be an awesome ep

Fluplaxio

YouTube has a small feature that asks you what you want your feed to look like every once in a while. It kind of sucks though because I’ve selected videos before and then it just takes that one genre of video and takes up every single video for a couple days. I watch a lot of different topics and I hate the one topic flood.

SxK

The thing with letting grok control the algorithm is that, if grok can do that, so could a user themselves because grok will not change the algorithm/code directly (would def just introduce bugs) but will interface with an existing API. So any decent coder could program a widget with a long modifiable tag list that also interfaces that API in an afternoon, that lets the user controll it, which would probably also work better since LLMs are not reliable. I don't see how a LLM could be of any help with a suggestion algorithm since they are designed to work generally on human input not specifically for sorting/ranking based on predetermined features, which is (probably) already handled by normal neural networks on these platforms

Pappanoodles

wait isn't more manually setting an algorithms taste to what you want one of the main things of bluesky. granted I have not used it much but from my brief experience a big part was that there was no algorithm but by every post was a recommend button which it would then use to make an algorithm for you instead of it being automatically done

Ni

I feel like Patreon's layout changes kinda demonstrate Atrioc's point - its feed used to only show you activity from the people you've subscribed to, and now the feed has a small section for people you're following and the rest of it is just stuff from top creators. I couldn't imagine paying someone on Patreon without having consumed their content elsewhere, so imo the change just makes your experience on Patreon worse, but it does feel like a clear intent to try to get people to spend more time on the site when the people they pay for aren't posting

pnott

That would require the LLM to run on the frontend, which it won't. The API for this will just be "send a message to grok" and everything else will be hidden behind that.

M C

10/10 DougDoug love this episode. Honestly, it’s great hearing someone being hopeful especially when even Big A’s thing was like “Elon can’t do this”. The spillover into other companies make sense. I do just have a general question about algorithms though. You guys remember Cambridge Analytica and how it eventually had to close because of the backlash? Why didn’t any of the algorithms get regulated during that? Like wasn’t that just everyday marketing now? Or am I fundamentally confused how this entire system that controls my life works?

Leo Maldonado

I gotta say I respect the hell out of Big A for actively trying to make sure he’s not stuck in an old opinion when it comes to Argentina. - per Atrioc

Salvadore Taranto

No more captions on the patreon extras, is that a patreon thing or was that one of the things y'all talked about in the Vox deal discussion? I'm a little weird and I like to read along sometimes, just curious

Lemonade Stand

I assume it's a Patreon thing, can take a look at why this may have changed. We may have accidentally toggled a setting or something too idk - Aiden

some guy

I want the damn bubble to pop already

Guy Fernando

“Accusing gentle Elon of a lie, that’s the last straw” 🤣🤣🤣

aleks

READ THE THE SINGAPORE STORY

Mike McG

Can you guys talk about the South Sudan situation? Aiden has tried to bring it up twice and has been cut off

Joseph Murphy

I recently just subscribe to your patreon but I can't find your Discord can someone help me

Sweath

You need to link your Discord to your Patreon account. Find the connected apps section in your Patreon settings to do it.

Schrodinger's Greed

If someone charged extra to not have a horrible algorithm, I wouldn't buy it. I would stop using the app.

Schrodinger's Greed

I feel like it's disingenuous to say that YouTube's chronological list for subscriptions is anything close to its actual algorithm. Can't speak to Twitter, though.

Schrodinger's Greed

Atrioc seems like a big fan of the Argentinian leader while also getting corrected on a lot of facts during the video.