Now that the fake demand is hopefully gone with Colette, we may have gluts in many parts of the supply chain. Good for inflation. Good for hyperscalers if they hold their horses on spending. Bad for semis and infra in general. Unless of course Elon picks up the tab. Possible.
QuantPM retweeted
Trusting and relying solely on what management says to make an investment decision rather than doing 360-degree research (surveying customers, studying competitors, speaking with suppliers, building your own forecasts) is not research. That is the classic definition of a cult. Many $TSLA investors are guilty of it.
QuantPM retweeted
Tim Cook would never cheat on his wife!
$AAPL
Ben Horowitz on what it would take to actually see an AI "fast takeoff" scenario, using a framework from Databricks CEO @alighodsi:
1. Each new model needs far fewer GPUs than the last
2. Each new model takes less time to train
3. Each new model is more accurate than the last
4. It keeps happening, again and again
"Zero of those four have been met... We're really talking about a risk that's entirely theoretical at this point."
@bhorowitz on @theallinpod
Ali Ghodsi of @Databricks on the four conditions that would have to hold for AI to truly improve itself:
"The next model requires less resources, less GPUs to train... The second condition: the next model takes less time to train as well. Third, accuracy of the model, the intelligence is increasing. And fourth, we can do the former three again and again and again... All at the same time, not just any of them."
"If you just mean that the software is writing itself, we're already there today. Like 90-some percent of the software in Databricks is written by AI. Does it matter if the last few percent is also written by AI? No."
"If you're getting these four conditions, then you might get a speed-up where the next model, let's say, takes half amount of time and half the resources, and it is more intelligent... By the way, I don't even know if that necessarily leads you to superintelligence."
"A lot of people, their definition is just, 'Hey, if I'm not even coding anymore and it's coding itself'... I think they're conflating, 'Hey, what's my value and is it scary for me?' Versus, 'Hey, that then means we'll get that superintelligence that in 2014 was hypothesized by Bostrom.'"
@alighodsi @martin_casado @sarah_wang
Aside from chips memory etc… what may become the biggest AI bottleneck is actually credit
(chart from Ben Evans).
$NVDA up 7% pre market mostly due to 70% growth in 2027 vs 45% street. This time marginal growth is coming from so called ACIE (ie all but hyperscalers). Those clients for the most part will need financial support. Ie Nvda will continue to increase c/p in the system.
QuantPM retweeted
Replying to @Kasparov63
When you have a room-temperature IQ, everything looks like superintelligence.
QuantPM retweeted
What I don’t entirely understand and honestly find pretty hilarious is that the $70B OpenAI ARR figure was released on Dev Day, on September 29
How did we go 9 days without anyone from OpenAI coming out and correcting the figure? For over a week now they’ve let people believe ARR was $20 billion higher than it really is. What does that say about OAI?
QuantPM retweeted
"We missed our numbers" -> unfortunate, disappointing, bearish
"We were pacing the frontier" -> noble, strategic, humanity-saving
QuantPM retweeted
The problem with working on any breakthrough using AI is that there's no guarantee the company running the AI won't steal it from you.
QuantPM retweeted
everyone getting sniped by the personal drama, but missed the more interesting unanswered question:
can these labs see all your work and scoop you when the stakes are "high enough"?
🤷🏻♀️
QuantPM retweeted
tristan buckmaster just published a statement that is, if accurate, one of the ugliest things i’ve read out of a frontier lab.
not because of the math. because of what happened after openai found out two people were about to beat them to the punch.
buckmaster and levent alpöge spent a year pushing the córdoba / martínez-zoroa program to smooth forcing. august 15 they got blowup for boussinesq and 3d euler. lean-verified august 22.
they were paying openai out of pocket and dumping every draft into private codex sessions.
the program is obscure. almost nobody else was on it. this was not “paste the clay problem into a chatbot.”
then this:
> thursday sept 3, buckmaster emails openai privately. rumors are flying. he is trying to stop a mess, not start one.
> they reply the same day: give us details so we can “avoid competing.” also, want free compute?
> sunday they suddenly need a call “at any point today.” sebastien bubeck gets on. levent is not invited.
> openai says an internal model has a ~100-page proof of forced navier–stokes blowup. fefferman options c and d. the exact route buckmaster and alpöge had quietly chosen.
> they show him a prompt and claim the model was “simply given the problem statement.” levent is told “very little human input.”
> that falls apart on the same call. a whole team had been on it. they started on the unforced problem, warmed the model up on easier equations including euler, and even the prompt they showed him was written by prompting codex.
> he asks when the first prompt went out. they stall. then agree: the past few days. after information about his work reached openai.
> he asks whether the model was trained on, or had access to, the private codex sessions where they put every draft of this project.
> answer: the model does not look up user data.
> he asks again. about training.
> no answer.
then it stops being a science story and becomes an hr story.
> offer 1: you post euler. we post navier–stokes the next day.
> offer 2: after you post euler, you alone write the navier–stokes paper, credit our model, and we cut levent.
> bubeck says twice he wants levent removed from authorship because he works at anthropic.
he declines both.
he says if they release it that way, he goes public.
the reply is not a scientific objection.
> “why would you ruin your career?”
> “if you don’t want me to be nice, then i don’t have to be nice.”
later, a text to levent proposing a one-on-one:
> “i don’t know if tristan is being fully rational right now.”
that is not how you treat a collaborator. that is how you split one.
openai got word two researchers were closing a path almost nobody else was on. they spun up a team on that exact path. they would not answer whether the researchers’ unpublished drafts in openai’s product were used. they tried to dictate the announcement order. they tried to remove a coauthor because he works at a competitor.
when that failed, someone reached for a career threat.
this is a lab finding out academics were about to publish, sprinting the same narrow program, refusing the data question, and trying to manage credit around a rival affiliation.
really slimy if even half of this holds.
what the fuck is happening.
holy shit.
tristan buckmaster is saying openai has a forced navier–stokes blowup via the córdoba / martínez-zoroa program.
and that it may have used drafts from his and levent alpöge’s private codex sessions.
that’s the córdoba / martínez-zoroa route. almost nobody else was on it.
he also says the first prompt went out after information about his and levent alpöge’s work reached openai. that “very little human input” was not true. that a whole team had already been set on easier problems, including euler. that even the prompt they showed him was written by prompting codex.
then he asked:
did the model have their private codex sessions.
no answer on training.
if any of that holds, openai didn’t just scoop a math result. they may have done it off two researchers’ unpublished drafts, then tried to cut the coauthor because he works at anthropic.
a millennium-problem-adjacent result just became employee drama before an ipo.
i don’t even know which part is crazier.
this is one of the slimiest things i’ve seen in this industry.
Congratulations on solving a beautiful problem. This is bringing back old memories ; I still remember first discussing this problem with Sergio Verdu when he was on sabbatical at Berkeley in the late 1990's . With a wink of an eye it's now 25 years later.
Some perspective on this problem from someone who has been around. This problem belongs to the class of problems called "multiuser or MIMO detection", pioneered by Sergio Verdu and received a lot of attention in the late 80's and early 90's, In the late 90's and early 2000's, there was a resurgence of interest in this problem via connections with random matrix theory. I myself have worked on a bunch of these problems, evaluating performance of different MIMO detectors. The problem that AI and @DimitrisPapai
have solved is the ultimate of these problems because the optimal MIMO detector is the joint ML detector. I am very happy to see it solved.
That being said, I have to say that MIMO detection as a whole has made little impact on how actual wireless communication systems are built. Because wireless communication systems, like all communication systems, use coding, while MIMO detection only talks about uncoded systems; the N information bits here in this problem are uncoded. It turns out that much lower SNR can be achieved with coding. Take this problem as an example, with coding, a constant SNR* = 1.32 is sufficient to support an information rate of 1 bit per antenna, in contrast to the growing SNR = 2 log N result just proved. That's one reason why although this problem is beautiful as a mathematical problem, it has not received the full attention of the information theory community even in those days of peak activity. Claude Shannon has already told us in his 1948 information theory paper: to achieve the ultimate limit of communication, one needs to do coding. Uncoded systems are sub-optimal. So SNR = 2 log N is only optimal among uncoded communication systems, but SNR* = 1.32 is optimal among all communication systems.
The problem that Claude Shannon solved in his 1948 paper was the capacity of point to point communication, The problem here is one example (so called point to point MIMO channel), and hence the optimal SNR* = 1.32 can already be computed by Shannon's 1948 theory (no AI needed). Most of the information theorists' attention in the hey days of wireless communications research has been to extend Shannon's theory to network information theory problems, such as the broadcast channel, the relay channel and the interference channels. In those problems , one ask what is the best information rate (or equivalently, the minimum SNR needed) that a network of users can communicate with each other. These problems have been proposed since the 60's and the 70's and to this day none have been solved in their full generality. I'd love to see AI take a crack at some of those open problems.
QuantPM retweeted
Replying to @AnnaEconomist
Reminded me of this from the CIA Manual "The Simple Sabotage Field Manual"
QuantPM retweeted
I’m not sure whether Dean Ball is confessing to a regulatory capture strategy or simply predicting this will happen (he now says the latter). Either way, the weaponization of regulatory uncertainty as a competitive tool should be completely unacceptable.
He argues there’s no need to ban Chinese open-source models — just direct agencies to issue soft-law warnings that create enough FUD so regulated enterprises back off. “It needn’t be that well justified.”
Wrong. Regulatory decisions should always be well justified and grounded in facts, logic, and evidence, not the deliberate exploitation of fear and uncertainty. Implementing a surreptitious policy through manufactured doubt — rather than strong and explicit justification — corrodes the rule of law and invites future abuse against anyone.
We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition. They have laid their cards on the table. It is time for the rest of Silicon Valley — the vast majority that still values open competition — to do the same.
Some observations on Kimi:
1. It's a very good model! I don't think its performance can be explained away by distillation or anything like that. In agentic coding sessions, it seems pretty much on par with the best public models of Q1 2026. In my fairly limited use, it also seemed very token hungry. It's not obvious to me that this model is actually that cheap to run.
2. I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks. To be clear, I *myself* might be fine with models presenting this level of marginal risk being open weight, but I am surprised that China is fine with it. I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). The other 25% or so is their lack of compute for customer inference (making China's open-weight strategy an unintended byproduct of US export controls) and the normal Chinese strategy of aggressive exports. For the companies, as opposed to the government, the decision to open source is partially ideological and partially because they are behind, and they know that very few people would pay for sub-frontier models from China.
3. Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. I suspect the reason they are is that they know open-weight models are effectively ungovernable, and they simply like the overall cloak of ungovernability open-weight models create over the whole of AI. It's not a bad strategy; it reminds me of James Scott's recounting of the hill people in "the art of not being governed." Still, in the end, open-weight models deter further AI capex.
4. One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a "public good" which will ultimately be provided by the state as a kind of "digital public infrastructure." This future strikes me as a dystopian hellscape, but I've never met an open-weight models advocate who doesn't ultimately concede this is where things end. You'd be surprised how many 'accelerationists' lobbied me, while I was in government, to support an eleven or twelve-figure federally funded data center so that startups could train models at a subsidy and then give them away for free. There was no other way for AI to progress, they said. Perhaps this is the logical end state of things. Nonetheless, I find myself surprised to see supposed accelerationists excited about such an outcome. I think many of them just don't know what they're doing. Many accelerationists do not view the creation and serving of frontier models as a legitimate business.
5. I would guess that the Trump Administration will at some point realize that their best strategy here would be to create large amounts of regulatory risk around the use of open-weight Chinese models. You don't need to "ban open source" (one of the dumber motifs of AI policy discussion). You just need to direct every agency to issue soft law that creates FUD. "A Federal Reserve Advisory Bulletin found that there may be backdoors in Chinese AI models." It needn't be that well justified. You just create enough regulatory risk that every regulated enterprise backs off. You probably don't want to create so much regulatory risk that you scare off the hyperscalers from serving Chinese models; this will just drive startups to sketchier providers. There's a happy middle ground here. I'd assume they will do some version of this.
6. It's probably true that open-weight models of this capability make the world a bit more dangerous, but not so much more that you'll really notice. At some point the models will be capable enough that you will notice. "A nonliving, invisible, dangerous, and infinitely self-replicating agent escaped from a Chinese lab," you say? Color me shocked.































