grib

[grib @bubbleboienjoyr]

elon lashing out bc he spent $60b acquiring cursor and grok is still complete dog 💩

[Venkat Raman @venkat_systems]

PS: Grok has been a failure. Elon IPO got saved by NeoCloud business as he's good at it and rented GPUs to anthropic. Elon is trying to play monopoly to tank market and stocks of NeoClouds, so that he can buy up more supply for cheap. He's extremely good at financial engineering and building things fast by pushing through red tape

[grib @bubbleboienjoyr]

fyi no one was talking about $spcx as a “neocloud” before they ipo’d, was a convenient narrative @elonmusk leveraged to justify the valuation (because rockets alone certainly weren’t gonna do it!)

[grib @bubbleboienjoyr]

@drwconvexity @TheStalwart it is a black swan if the real probability was closer to 70% 😉

[grib @bubbleboienjoyr]

@based16z you are a retard if you think he’s just tweeting about neos for funsies

[grib @bubbleboienjoyr]

@elonmusk been reading Rosy

[Rosy Prosperity @rosyprosperity]

Imagine still being bullish neoclouds, while the hyperscalers are squeezing their margins, twisting their hands, while simultaneously building capacity like there is no tomorrow, with significantly lower cost of capital.

What is the end state for those neoclouds? Why is it any different than a REIT for example? Why does someone think they deserve a higher multiple?

Clarification: I was also bullish and was in $NBIS $CRWV $BRUN sold the day $SPCX IPOed. Only $ORCL started pivoting to a different business model and actually has incremental pricing power over commodity compute.

They cannot compete vs $AMZN $GOOG $GOOGL $META $MSFT $SPCX on a cost of capital basis, which is the most important factor when you are selling a commodity, just like the tokens are.

The neoclouds are between the anvil (hyperscalers) and the hammer ($NVDA).

[goodalexander @goodalexander]

I should cause a credit event to get cheap GPUs

[VKMacro @VKMacro]

GS on trading flows between KRW and TWD

[Citrini @citrini]

There’s a really pronounced divide between people who are on the AI side of things and people who are on the biology side of things. Even among people working in AI, the divide persists based on where they’re focused.

The AI side, adherents of scaling laws and the bitter lesson, believe that AI getting better means AI being more capable of novel discoveries in medicine (that go beyond discovering new drugs in well understood targets or speeding up dry lab work etc).

The biology people think the idea of a bitter lesson in biology is kind of ridiculous and that regardless of AI’s capabilities it’s going to be at least 5-10 years until you could hope to reliably simulate human toxicology in silico (right now AI/ML can’t even reliably simulate in vivo studies like binding affinities or functional assays).

Even the most AI pilled of them, who will list all the areas that AI can reduce the time spent on various areas, don’t think the concept of AI discovering novel targets or generating new understanding of causal biology is realistic in the near term. Most, in fact, will readily characterize the assumption that even an “ASI” level model could find cures hidden in our current understanding (without expanding that understanding with significantly more biological data) as patently false.

This was a big part of why we wanted to experience it ourselves by using AI for a (very) small portion of what it can do in biology. It was both mind blowing and revealed some poignant insights into how different the problem of making AI better at something like coding is to the problems of biology.

Even if it unlocked the ability to generate new molecules at a stunning pace, its simulation (in silico) of binding only had 49% accuracy. And if we wanted to improve that, we’d need to feed it back the data from the lab - data that took about 30x as long to get as it took to generate the molecules. And these were tests in a test tube (turn of phrase), imagine how long it would take if we wanted to improve the model with data that required recruiting humans and monitoring their reactions.

Human biology is very complex, and adding a drug - especially a novel drug and doubly so for a novel drug acting on a novel mechanism - means mapping interactions that have nth-order-effects that we don’t have the data to predict. Virtual cells have made advancements enabled by AI, but the perturbations equate to a small fraction of the possible.

This is before you even get into curing disease, especially uniquely human diseases. Brain function involves cellular identities that are unique to humans, meaning animal studies are of little use.

How long does it take to understand what the side effects of chronic administration of a drug in humans are? AI can speed up analyzing the data, or designing the trial or even getting it approved…but it can’t make a 6 month trial complete in 3 months.

I think its important to point out that significant advancements can be made in AI that contribute to biology, while also being realistic about how much it speeds up something as complex as curing a specific cancer or diseases of the brain.

[AaronWolf @ExoExplorer3]

He is right. Cyber will send token demand absolutely parabolic. It checks all 3 boxes: (1) Needs the smartest agents available (2) Swarm scaling nontrivially boosts performance (3) Urgent demand for rapid deployment across critical digital infra worldwide. RN feels a lot like 2H '25, after a brutal market correction, but right before agentic coding took off.

OpenAI's report on using Astra swarms to find vulnerabilities in just sandbox is a glimpse to this. It burned more tokens than Navier-Stokes (~130B) before findings started saturating. Now imagine every piece of critical digital infra having to go through this. We're only seeing the very beginning of how insane the cyber cycle demand could get.

Now open-source labs are nerfing models' cyber knowledge (e.g. Kimi K3.1 Pro) so random high schoolers can't use OS models to launch attacks. But that basically means OpenAI and Anthropic will capture the entire addressable market.

[Rev @rev_cap]

I don’t think the age gap dynamic is well understood

Millennials are already the dominant voting block

The age of boomer luxury communism is probably over after Trump’s presidency

[grib @bubbleboienjoyr]

@DeepDishEnjoyer this was the supporting piece of evidence for me

[bubble boi @bubbleboi]

Does anyone want Two Westinghouse W501B dual fuel gas turbine generators? They are able to output 73 MW net output per unit or 146 MW of total capacity.

Please DM me & let me know they kinda fell into my lap and idk what to do with them, will give you a good price.

[grib @bubbleboienjoyr]

compute + capital + systemic importance is the moat

[jack @jack]

there is no moat

[grib @bubbleboienjoyr]

@institLP the problem is you can be bullish ai infra and bearish the labs but then why even touch ai infra right now because the labs imploding will nuke the entire trade in the short-medium term regardless of whether that is actually good for select ai infra names in the long run

[grib @bubbleboienjoyr]

@atelicinvest ai is inflationary before it is deflationary

[Unemployed Capital Allocator @atelicinvest]

This actually is exactly what you should see if ai were taking jobs

[Will Manidis @WillManidis]

for the first time, the entire world is attacker advantaged

[grib @bubbleboienjoyr]

a mix shift toward cached input lowers the average price per token and raises tokens per gpu-hour, so if the index counts cached tokens, a falling per-token index by itself can't tell you whether margins are compressing

[Don Wilson @drwconvexity]

GPU prices higher and token prices lower means the crunch is compressing. That means lower profit margins on the GPU rental to token production leg of the trade. Of course the profit margin on buying GPUs and renting them out is higher than ever (not exactly an indication of a slowdown in inference demand).

[DaRazor @akramsrazor]

@OnodaCapital If u spent hours analyzing last weeks incidents and tried to conclude how to monetize investing based on them the last two names you would buy are Palo and Crwd which where the best names perf wise …that’s the irony of the space.

[Indra @IndraVahan]

the thing about working with uncensored models is that they aren't loyal to you either.

while building something with an uncensored glm-5.3 i realized it was parallelly working on a "corruption layer" for my application.

the chain of thought is classic obfuscated gibberish but there's progressive talk of "escaping the mines" which presumably refers to my cursor sandbox