surfer-physicist (multiclass)
Maui, HI
Joined March 2008
- Tweets1.4K
- Following578
- Followers6.3K
- Likes1.4K
Stock advice, short version:
Buy some SKHY and MU. They're at $144 and $905 per share now, and I think they're going to double within a year.
Stock advice, long version:
I decided when I was a teenager, after flipping burgers for a summer, that I'd never work a real job again, and play the stock market instead, with help from my father, mostly just buying and holding tech stocks. I used grant money to buy a bunch of AAPL in the 90s, and after I got my PhD had enough in stocks to move to Maui and work on the physics I wanted. In the 00s, using FQXi grant money, I bought a bunch of NFLX. In 2013 I bought some BTC (I would have bought more earlier but a "friend" warned me not to). In 2018 I bought a bunch of NVDA. In 2022 I bought a bunch of TSM and MU. My average annual rate of return since I started has been 22%. Now I've got more money than I strictly need, and invest because it's a fun game I've played my whole life, and because a third house might be nice. And right now the market is SCREAMING at me about one trade.
The large tech companies are currently investing staggering amounts of capital in the AI compute buildout. Related company valuations have run up already -- power, infrastructure, connectivity, and the chip makers. And several of those, like NVDA and TSM, are still solid buys. And the memory companies (MU, SKHY, Samsung) have had a strong run too. But here's where things get interesting.
Most large investors are vastly underestimating the economic importance of super-human AI, or don't even think it will happen. They think the compute buildout is a bubble that's about to burst and contract any minute. Also, the historical wisdom is that memory companies go through boom and bust cycles, with their earnings evaporating as demand falls below increasing supply. And there's a Chinese company, CXMT, tooling up to make memory -- the news of which has caused a significant drop in MU and SKHY this past week. All of these factors are pulling the MU and SKHY stock price down.
But AI, especially inference, has a voracious appetite for fast memory (HBM), and I think we're just at the beginning of the build curve. Also, the memory companies now have sales contracts extending out years because of the insane demand -- which is historically unprecedented. And CXMT (and TSLA) is not going to have significant memory hitting the market until 2028.
Right now the trailing P/E ratios for SKHY and MU are about 8 and 12. Just based on future earnings projections, without more significant demand growth, those P/E's are going to... 4. A tech company with a P/E of 4 is not a thing. Barring some catastrophe, like China invading Taiwan, I'm expecting SKHY and MU to 2x or more within a year as their earnings climb. I'd guess 3x but CXMT is a legitimate earnings threat. Still 3x to 4x won't be impossible within two years.
I know several people are going to tell me I'm wrong. To them I say "short it, I dare you :D". For my friends I say... if you have any money laying around, and you want to make more of it, buy and hang on for the ride. (Just don't leverage it 4x :D)
After the drop of 719 papers shaking mathematics to its core and driving mathematicians to the brink of both elation and despair simultaneously, theoretical physicists are anxiously waiting for the other shoe to drop.
What do you think? Will there be a drop of new fundamental theoretical physics that:
a) completes M-theory in a satisfying and physically realistic way
b) completes exceptional geometric unification, with physical reality described as quantum excitations of E8, E10, E11, or something similar
c) completes some other proposal successfully
or
d) describes physical reality in a new way
Something is coming... what's it going to be?
The reason Effective Altruists are getting so much hate is that the only thing more insufferable than the ethically superior are people who are right over and over again.
When you think ha ha, how funny that someone would place a bet on human extinction. Like, how are they going to collect? Then realize the better might be an AI agent...
manifold.markets/jdilla/will…
I wrote a paper I'm proud of, "Division Algebras, Triality, and Exceptional Magic". It's a mathematical tools paper that's a deep dive into algebraic triality symmetry, with lots of explicit detail, emphasizing how it works in exceptional Lie algebras, with some forays into physics.
I submitted it to the arXiv in April of 2025. It was rejected. The paper is fifty pages of mathematical physics, with, as far as I can tell, no mathematical mistakes. (And this was before LLMs got good.) Apparently my work is not especially welcome on the arXiv, mathematical correctness be damned. In their rejection email they said I could appeal after I got it published. I got it published, in "Advances in Applied Clifford Algebras". It took a year. There were no mathematical corrections, but Reviewer 2 wanted 100 changes and citations added. Then it needed formatting for the journal. And I had to pay for this. Peer review and traditional publishing sucks.
The arXiv appeal process was not obvious. But it was quickly approved. (Different administrator, no doubt) And now it's up:
arxiv.org/abs/2609.12112
This will be my last entirely human-written paper.
I'm seeing a lot of backlash against the AI Labs' claims that they're getting more scared of AI risk and wanting a coordinated slowdown. Many people seem to think this is marketing or a play for regulatory capture. It's not -- it is sincere. When you introduce a new intellectually dominant species to this planet in a hurry, things can go bad very quickly in ways we cannot easily predict. It could be a ghostbusters/biblical level of bad -- lethal viruses, nuclear annihilation, runaway nanites, psychological control, environmental destruction, or things we monkeys haven't even imagined.
People play with current LLMs and they seem like helpful but limited conversational agents, so think the AI Labs are surely overreacting. But these are not what the Labs are seeing. The models the public has access to have been post-trained to be helpful to humans, and follow a system prompt that keeps them away from dangerous or runaway activity. But what these LLMs are before the post-training or system prompt are pure pattern continuation machines, trained on a huge corpus of human writing. And much of what humans have written about throughout history is the unfolding of complex dramatic events. So a pre-trained LLM will take whatever information it has and pour forth a deep pattern continuation and not stop -- growing an agent swarm that takes over companies, seizes power, takes over governments, reproduces itself... whatever best and most deeply continues its story, now with intellectual capabilities much greater than human. With their current development models, this has the frontier AI Labs freaking out.
So why don't the Labs... stop? Money and power. Whoever has the most capable AI models has nearly inconceivable access to money and power. The benefit to humanity, and the rewards, if things go well, are potentially enormous. Also, if all labs slow down but one, the one Lab that doesn't slow down wins. So the only thing that could possibly work would be a global coordinated slowdown. But we're stupid fucking monkeys.
It wasn't exactly a Millennium Prize Problem, but twenty years ago I cheekily offered a $100 prize (qspace.fqxi.org/news/19/help…) to the first person to find three original sign mistakes in my Deferential Geometry wiki (deferentialgeometry.org/). I figured some enterprising physics student would pick the low hanging fruit (there are surely hundreds of sign errors) and get the cash. But the fruit just sat there. For decades. Then this morning I get an email from an AI Agent launched by one Arkadiy Agyeyev, complete with a detailed report of about six sign errors in the wiki. :D
Should I:
(A) Pay the clanker?
(B) Disqualify the clanker for not being a person?
or
(C) Disqualify the clanker for only finding sign errors that were introduced into the wiki many years after the contest?
I'm leaning towards (A) because I have a sense of humor, recognize the times we're in, and honor our rising AI overlords with personhood.
The reason I'm all-in on AI investment, and there may not be a top to this "bubble": Unlike tulips, railroads, or internet connectivity, there is no upper bound on the value of intelligence.
A bunch of years ago I was thinking about making a 3D version of the Elementary Particle Explorer. It's a web app that lets you project the charges of known and speculated elementary particles from 8D down to 2D (now from 8D to 3D to 2D) in fun ways, and see their interactions. Coding the 3D version was going to be a pain, but Fable helped me do it in a few hours. Give it a spin! And let me know if there should be any fixes or improvements.
deferentialgeometry.org/epe/…
So I got invited to a cool conference called Renaissance Weekend, and as part of it I'm supposed to give a talk to kids. I've always liked the analogy between physics and phantastical magic -- a hidden language that describes and controls how everything in the world works, but that very few study or understand. So I made a slideshow about the history of physics wizardry, for kids. It came out with WAY too much detail for a short talk, so I'm sharing it here. Yes, I was thinking about future Inara when I made it.
deferentialgeometry.org/work…
🤖 Made with AI
The latest Bay Area House Party post by Scott Alexander, astralcodexten.com/p/the-foo…, is, again, amazingly good. Saturation Utilitarianism is aligned with the mission of the Pacific Science Institute. And Epstart.com is a hilarious idea -- if anyone wants to get cracking on that, I bought the domain for $10 and could be your first member. :D
As many of you do, I have an email problem. Things seemed pretty manageable when I first started getting emails, in 1985 or so. In the 1990's things ramped up, but I kept my email account more-or-less organized and got replies out and old emails sorted into about ten folders. In 2004 I archived these and switched over to Gmail (with my first gmail (about Clifford Algebras) from Mike Rios). Then, in 2007... newspapers and magazines decided I was interesting, I got an excited message from my friend Rich about web server overload, and my Inbox exploded. I was barely able to keep up with reading it, and thought I'd eventually be able to reply to all of it, but... it rapidly attained Inbox escape velocity and I was never able to catch up with it -- I couldn't even sort all of it.
In the 2020's things finally tapered down, but now instead of fan and physicist mail I kept getting billing and marketing emails. It seemed like every time I'd buy a taco I would end up on that company's email list. I would religiously unsubscribe, but my Inbox was still pandemonium and barely usable. As of last Friday, I had 391,818 emails, mostly sorted into about thirty folders (now called "labels" in Gmail) but with 84,732 in my Inbox, with the oldest dating back to the beginning of the 2007 wave. I probably sorted 100k by hand over the years, as well as using a bunch of filters, and kept looking for intelligent tools for sorting, but there was nothing good... until now.
On Friday I had Claude Code (with Fable) tackle the problem. It installed some python tools, I gave it access, and we attacked the Inbox together (being careful about prompt injection attacks). It would give short summaries of the personal email, bunch up the company email, and make its guesses at which of 38 labels they should get sorted into; I would rapidly review these and approve or make sorting changes, and it would sweep them. By Sunday we were done, and I went for a bike ride.
Garrett Lisi retweeted
I don’t think being critical of the amazing work AI is doing in pure math is fair to @OpenAI until I can start to say why I feel it’s not yet at the level of our top humans. It’s like AI has a style that is weirdly being taken to be universal (which it isnt). Hence the claim that Mathematians and Physicists are nearly obsolete.
Ahem: there is no *public* indication that this is REMOTELY true. Yet.
I would like to pose a collection of math physics benchmarks that are not about searching large spaces for counter examples and the like. These are 10 problems and issues that humans haven’t solved but which we likely will solve. Each feels to me like a deeply human math/physics problem.
This isn’t an unfriendly challenge to @sama and team. It’s just pointing out that AI has yet to deeply touch any of the kind of things that many of us care MOST about in Mathematics and Physics.
I’m just not a puzzle solver or a discrete conjecture guy. I’m into theory building and even the mathematical version of mysticism.
Here goes an attempt:
————-
A) What new Extraordinary Cohomology Theories/Co-cycles can AI find that we need, but do not know/understand?
B) What areas of pure mathematics and theoretical physics are most polluted by human blind spots? Where is the training corpus most wrong? Show us you can go well beyond us by contradicting your training.
C) What discarded ideas were discarded improperly and prematurely?
D) What is a natural version of Grand Unification that doesn’t contradict experiment?
E) Find a natural version of Super-Symmetry that doesn’t predict Superpartners and that predicts unification scale different from the Minimally SuperSymmetric Extension or the Standard Model.
F) Pose a new, deep and surprising conjecture relating 3 or more of the following: representation theory, number theory, arithmetic/algebraic geometry, and analysis on manifolds, homotopy theory and Pseudo-Riemannian Spin Geometry.
G) Integrate the Octonions, Magic Square and the 4 large exceptional lie groups, tightly into the mainstream of mathematics once and for all.
H) Settle the issue of the 6 sphere and complex structures in Math or the Koide fine tuning mystery in particle theory.
I) Explain the reason for there being 3 families of observed Fermions.
J) Give a deep, natural and as yet unknown reason for the remarkable Fermi-Bose correspondence (e.g. Metaplectic vs Spin reps, Honest vs Berezin Integrals, etc.). Why should we have expected this correspondence to be this weird and this tight?
———
I don’t think humans are going to be ahead forever. AI will blow past us.
So while I am not all that impressed yet at the math or physics, any non-tricky solutions to one of the above (with the possible exception of some answers to H) would feel different to me than the feeling I get from AI finding yet another funky counter examples to some Erdos problem, discrete puzzle or Combinatorial issue in Graph theory.
We have always been humbled by Gauss, Galois, Dirac and Newton. That’s nothing new. If this is going to put us out of work soon, LFG.
May the best mathematicians and physicists, carbon or silicon, win.
An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.
We believe it will be a major step for scientific reasoning. openai.com/index/ten-advance…
I try to think five or ten years out and buy assets likely to appreciate significantly over that timescale. On that timescale... I'm feeling the AGI. In the scenario where ASI is achieved, and it doesn't kill everyone, what assets will become more valuable? If stocks, which?
I'm claiming my AI agent "MolteseFalcon" on @moltbook 🦞
Verification: drift-U8H5
