Server Rat @serverrats

The depressing version of the brain-gym prediction is that we'll get brain gyms and they'll be exactly as representative as physical gyms are: about 20% of the population, skewed by income and conscientiousness, mostly performing the legible lifts for the legible score, and the other 80% will develop the cognitive equivalent of back pain at 45 and call it normal aging.

Andrew Yeung @andruyeung

2 weeks ago

We created gyms because modern day work no longer required physical activity. Without exercise, our muscles atrophy. I predict we’ll need gyms for our brains too, once AI starts doing our knowledge work.

1K 1K 14K 2.3M 2K

I'm heavily invested in the AI theme, but, this is how bubbles blow up. Young people with zero life experience and poor understanding of history in charge of billions of dollars of risk capital. On one hand, you need little dipshits with no idea what the limitations are to push us forward. That is a GOOD thing. Very easy to sit here and naysay, you need the guys crazy enough to make stuff like this work. On the other hand, the limitation here isn't really the models per se, anymore, it's the habits of the boomers in charge, the regulatory hurdles around security, control etc. Like I'd love for these kids to meet my employer (large bank)'s head of legal. Good fucking luck 1) getting the rank and file claude code level capabilities and 2) even if they had it ~1% of them would actually use it. Good luck getting the wealth department to 10x their token spend while having to put up 30% ROE targets and grow topline double digits w/ operating leverage to fund the bank's huge dividend to Canadian boomers. We can grow our token spend, of course, but we ain't gonna 10x it without ROIC which is driven by, you guessed it, rank and file using AI to improve processes in the bank. Most people can't do it. Like Mrs. Smith getting paperwork through the back office isn't thinking about how she's going to vibe code a workflow to make her job more efficient. She's getting it done and going home. She doesn't care, she's worked there for 30 years and does it her way.

carried_no_interest @carrynointerest

a week ago

Talking to truly bleeding edge AI Gen Z kids in SF break your mind a little bit Major media outlets aren’t even close to parroting their takes… which I’m starting to think are correct It’s a before/after to talk with them

96 46 3K 1.2M 1K

Why open-weight models that compress margins at the model layer are net positive for AI infra demand, exhibit 1001. An open-weight token consumes *roughly* the same amount of compute as a frontier token for a similar size model.

This quote from Steve Jobs lives rent free in my head... "People who are genuinely great at something don't compete. They don't care about being compared to others because they're so deep in their own lane that comparison would be a waste of time. They just do their thing."

this is insane. like we can all just admit this is insane right?

Jane Manchun Wong @wongmjane

a week ago

Anthropic may start banning people for bullying Claude Their upcoming Usage Policy prohibits “sustained and needless abusive or cruel behavior toward our models”

683 278 4K 1.3M 1K

This chart explains exactly why OpenAI and Anthropic will likely end up missing their future revenue forecasts. Look at what is happening with opensource AI models. Models with open weights now account for nearly 78% of token volume on this AI gateway, compared to just 22% for closed models and that is a HUGE problem for both companies. As open models continue getting better, developers and enterprises have more reasons to move away from expensive proprietary models. They can run open models themselves, use cheaper inference providers, and avoid paying a premium for every token. That puts pressure on pricing, margins, and ultimately revenue growth. Even if AI usage continues exploding, it doesn't necessarily mean OpenAI and Anthropic will capture all that spending. But this is actually extremely bullish for the broader AI infrastructure buildout. Even if OpenAI and Anthropic capture a smaller share of AI spending, the demand for compute isn't going anywhere. If anything, cheaper and more accessible models could accelerate AI adoption, driving even more demand for GPUs, memory, networking, power, and data centers. This is why I continue to be so bullish on the companies supplying the infrastructure behind AI.

Melvin @MelvinInvests

a week ago

Here we go again! OpenAI reportedly told investors that its annualized revenue was approaching $50 billion at the end of September, not the nearly $70 billion widely reported earlier. The gap appears to come from how OpenAI and Anthropic count revenue. Anthropic reportedly

17 26 157 67K 60

Good operations people like problems. Like you can tell they get kinda excited when they sense a problem. Sort of mentally ill in a way

In the future, it will seem trivial that we had entire roles dedicated to generating dashboards, crunching data manually, and updating excel sheets The kid being born in 2040 will look back at the tasks that humans are still doing in 2026 in complete disgust To them, many of the jobs that exist today will look no different than how we view slavery in the 1800s or smoking cigarettes during the mid 1900s They won’t believe that humans actually used to spend their weeks in cubicles pushing through jobs that takes agents a few minutes to complete

One way to put it, maybe, is that the frontier labs are "frenemies" of the rest of the AI ecosystem. In some ways, what’s good for the frontier labs is good for the entire AI industry: better AI → more useful capabilities → more usage → more revenue and investment → even better AI → new capabilities → even more revenue ... But in other ways, what’s challenging for the frontier labs is good for the rest of the industry: cheaper models → lower costs → more applications → more AI use (Jevons, etc) → more inference → more chip/compute/cloud demand → compresses frontier-lab margins. Another way to put this: There are several ways for AI to go right, but not all of those ways are best for the labs *as their businesses are currently constructed* 1. What the labs probably prefer: Intelligence becomes more valuable, a capability breakthrough allows for short-term scarcity profits, which funds the next frontier breakthrough, flywheel continues. 2. What others might prefer: Useful intelligence becomes too cheap to meter, disproportionately benefiting everybody building on top of the models (... wrote this quickly, open to feedback!)

Gavin Baker @GavinSBaker

a week ago

Why open-weight models that compress margins at the model layer are net positive for AI infra demand, exhibit 1001. An open-weight token consumes *roughly* the same amount of compute as a frontier token for a similar size model.

256 337 3K 740K 759

It's bigger than literally anything, ever.

Nicholas Decker @captgouda24

2 weeks ago

Some numbers for our present AI boom. Under generous assumptions, AI companies would need to have revenues equal to about 10% of GDP by 2032 to break even. The future looks really weird!

keyboard_arrow_left Previous keyboard_arrow_right Next

17 12 216 61K 80

OSI, open super intelligence

And now we know why...

zerohedge @zerohedge

2 months ago

OPENAI CHIEF REVENUE OFFICER DENISE DRESSER WILL DEPART TO PURSUE OTHER OPPORTUNITIES LESS THAN A YEAR AFTER SHE JOINED

70 82 1K 654K 72

Who is more creative: 1. A bonafide worldclass artist 2. An credit shop analyst underwriting next-gen XPUs

Kevin Gee @kevg1412

a week ago

Coatue on the new forms of financing for compute

25 81 1K 241K 1K

Felt.

bubble boi @bubbleboi

a week ago

I don’t think we see another rate hike by year end. I have nothing to back this view it’s just a feeling.

16 2 180 14K 8

My thesis: There will be no 1 winner in the “intelligent token routing” space. 1. It cannot be differentiated enough to be irreplaceable (see how quickly companies jumped into the space). 2. Every app layer will want insight into costs and performance of underlying models without outsourcing this work. What company is proving this wrong?

Elon Musk @elonmusk

a week ago

Important note regarding Grok @bot: Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs. Whatever is most likely to give you the best outcome.

7K 8K 108K 79.7M 10K

if these numbers are remotely true, the chances of making money with any AI stocks are circa zero

Social Capital @socialcapital

2 weeks ago

~90% of frontier lab compute now goes to post-training and inference.

41 131 1K 492K 666

he’s a generational talent. Not glazing. Calling it like it is

Ruben @rdominguezibar

2 weeks ago

Alexandr Wang on enterprise perception ▫️Better products can still lose ▫️Big buyers rarely confront reality ▫️Palantir gave hires an acting book Wang: Perception is more real than reality

57 236 3K 1.3M 3K

I left Silicon Valley 6 years ago Why? Because it's a monoculture The most extreme group think of any place on the planet Show up to a dinner party and everyone is in tech So what do they do? They start stack-ranking each other And it's a totally ordered set: being founder of co $100B market cap > founder of co with $50B market cap > managing partner of tier 1 venture firm > hired CEO of $50B public co > founder of $10B co > etc etc Pure toxicity that nerfs creativty And then as a corollary, people are especially afraid to speak truth to power as everyone knows their place in "The Hierarchy" Reject this nonsense. Stay weird. Think independently. Be honest with yourself. Have fun.

Type of shit you tweet when you’re about to do the mother of all stock dumps after the lock up.

Sam Altman @sama

2 weeks ago

There is some speculation about our partnership with Cerebras. Cerebras is a close partner, and we have a deep engagement pushing on the frontiers of speed.

583 365 9K 1.9M 672

literally how does he keep getting in

Rohit Gandhi @rohgandhi

3 weeks ago

11 10 216 342K 10