routers eat the world; consilient software observer; crafting an alchemist's paradise. economic meta-learner. DMs open.
palo alto
- Tweets 8,919
- Following 641
- Followers 5,930
- Likes 8,224
Pinned Tweet
Solving the science of asset selection in a future (or indeed the present) where every company is a "Context Acquisition Company" is the real frontier.
I love that everyone is getting around to the idea that the secrets (scarce context) currently illegible to/hidden from computers (human or machine) are everything. Now the next leap for people to make is that the science of sourcing, selecting, and monopolizing that context (really THE ASSETS that produce it) is everything. If AI progress is a function of compute and data (most algorithmic progress is really just data progress; h/t @BerenMillidge, @_kevinlu, @mentalgeorge, @GarrettLord, etc.), then every company is going to have a context desk just like they will (or already do) have a compute desk. The difference is, CONTEXT IS NOT FUNGIBLE. Most context (both that exists right now and that will be created in the future) will be completely commodity beta. Winning will be about getting to and instrumenting the right asset (context production factory) first. And yes, there are right and wrong answers. To do this kind of asset selection well requires an extremely scarce meta-capability: the ability to coordinate the right kind of access and the right kind capital at the right time.
These assets (and the secrets within them) are structurally difficult to access, evaluate and instrument. They are not floating around in banked processes, to be frictionlessly purchased on listed exchanges, or willingly coming through Mercor or Handshake's expert portal. (Yes, a context production asset can be (very often is) a single person or collection of people.)
When @WillManidis talks about a Deal Guy Yuga, what he means is that there are people who have deeply internalized the fact that at the limit, in a world of infinite intelligence, access to/monopoly on the right permissioned data streams is all that matters. Getting yourself to a position (meta-access, meta-capital) where you have the ROFR on those permissioned data streams, means being a generational Deal Guy. This is a very different and specific kind of "Deal Guy" though.
Knowing which asset(s) are going to give you the right context to create, compound, and commercialize the best vertical world model now and into the future is the new form of security analysis. But the triple-exceptional combination of domain expertise, meta-access, and technical ability that’s required to execute this new security analysis effectively is scarcer than the talent at quant firms, YC combined, and dare I say, the labs, combined. Palantir understood this and it's why they focused on getting root-access (or something close) to the "highest-status" institutions, and the data streams they produce, first. If you have the talent that can get access to and create value within those institutions, everything else should be a forgone conclusion.
If you want examples of the teams that (I believe) actually understand this new science of asset selection and long term value capture in a world of infinite intelligence, study Long Lake and @formationbio. They know and have known that it's all about being able to get the right asset (context), in the right market, with the right team (machine and human) first. These two companies are very far ahead on the scientific frontier of context acquisition.
GC backed Long Lake last year. Do you think it’s a coincidence that Long Lake chose to work with General Catalyst? My bet is that Long Lake knew they wanted to acquire Amex GBT before they partnered with GC, and that they partnered with GC because Ken Chenault (the ex-CEO of Amex) is General Catalyst’s Chairman. That gave them the right access at the right time to a very valuable context asset (Amex Global Business Travel)
A superhuman vertical-specific Elon operating every company means market leading monopolies in every single slice of the unstructured economy. The thing is you have to build this superhuman Elon while flying the plane. You can't build this superhuman Elon without the very specific context that operating specific assets in the real world gives you. In fact, there's only one stream of context that was able to produce human Elon! Knowing which context stream is likely to do the same a priori is so extremely difficult, but probably possible.
I’ll let you intuit why Amex GBT is both most likely to be the market leading monopoly if it were operated by the superhuman Elon of business travel and why it’s also the most likely to produce the context to build that superhuman Elon.
The labs of course are very large acquirers of context at present and I think they will continue to play and improve their capabilities here. Through their deplyoment companies, they have already chosen the PE funds that they deem to be the best Context Acquisition Funds. Through in-house deployment focus on Life Sciences they have chosen the vertical they see as containing the most valuable context producing assets. They will acquire very seemingly unrelated companies and will acquihire very interesting people just to get tokens, they will create a Context Acquisition Fund of Funds. But it's not a foregone conclusion that they become the best performing context acquisition companies. Or that they even view it this way. And that presents an opportunity for anyone that does.
The Context Acquisition Company (CAC).
We are a holding company acquiring services firms for their tokens in order to build domain specific agents/ models to deploy into our platform businesses and beyond.
There's $1T hidden in the computers. We're gonna get it out.
5
8
112
53,856
Interviewing can give you access to people you’d otherwise have no business being around while lowering the opportunity cost of following your curiosity. Both make it easier to become someone interesting enough to be interviewed!
when did a ton of ppl want to effectively be interviewers (podcasters) instead of being interesting enough to be interviewed?
the latter is a much more ambitious endeavor than the former. if anything you should become an interviewer after you’ve done something interesting.
same thing applies to being an investor (i.e. you’ll be much more respected as an investor if you’ve been a founder). before someone comes at me, there are very obvious exceptions to this rule.
5
500
I don't think this causes a run, just an upward pressure on deposit rates and therefore, all else equal / as a first order effect, lower bank margins. There's been extensive research on how reduced search / transaction costs affect deposit rates (hint: it's not great). But! Lower search / transaction costs don't magically take deposits out of the system.
Lots of people think agents will be like ozempic for human decision-making (i.e help people resist temptation / become more disciplined and "rational"). I suspect they'll instead make us more effective at actuating on our desires as they currently exist, not as we wish them to be.
If someone wants to spend more on the things they enjoy, an agent that helps them discover and access more of those things likely increases their spending.
Agents will also help us enumerate latent desire (sense and act on desire we didn't even know we had). Finding better uses for money generally increases the desire to spend it.
Increased consumer spending = more business revenue. Whether businesses retain that cash or pay employees and suppliers, those payments generally move deposits between accounts. Unclear the longer-term effects on deposits and bank margins. Lots to consider here.
But the broader corollary is more interesting IMO: the easier agents make it to satisfy our desires, the more consequential it becomes how those desires are formed.
Therefore, the most important work of our time is to find ways to increase the QUALITY of our desires: to make our desires a function of epistemic novelty rather than sensory novelty.
For this to happen, new institutions and institutional incentives are required. And there's not enough people doing meaningful work here. To me, this is a concrete alignment problem. An agent could become extraordinarily good at satisfying our preferences while the institutions shaping those preferences remain terrible.
Fascinating.
Chief Economist at Apollo: agents could cause a bank run by sweeping household cash into accounts paying 3-5% instead of the 0.1% national average, causing banks to lose a large share of their cheap deposits.
6
2
46
5,630
Economics of a neolab = Sum of The Parts Analysis of how much Zuck would pay in cash and stock for each researcher.
The economics of a Neolab.
A neolab is loosely defined as a startup of AI researchers who raises a lot of money pre-production to be able to finance GPU compute to take on a large AI problem.
To buy 1000 GB300s or ~14 NVL72 racks will set you back $125-150M for 3yrs with 15-30% upfront. That’s about ~2-2.5MW.
Thats about enough to do 10^25 flops a quarter and get to a GPT-4 level model which is 1-2 OOMs off frontier for pretraining.
If you post-train on a great open source model, you have a better chance of getting to frontier. The risks are a) you need to spend millions on RL environments too and b) being lapped by another model release while being tied to a base model.
For this to payback, you need to give your customers a better and ideally cheaper inference service than a base model and serve them for long enough to recoup your large investment. Even at 50% margin on inference, to recoup $10M in training means serving ~10T tokens (!) if you price like Fable / Astra given a standard cache read / input / output split ($2/M blended). And you have to justify being better than a release like Opus 5.5 which is even cheaper. Often, you end up charging your customers a huge premium in terms of platform fees and compute fees on top of pure inference.
Meanwhile, every hour you’re not utilizing your GPUs you are burning money so you typically resell this compute back to a broker or run inference for open models / resell spot instances. At below a ~60% utilization on spot, you will still lose money.
Add to that insane cost of talent.
So what can you do with the compute?
- Not play the model game at all.
- Play an entirely different model game (Jev, World Labs) that if big labs played, would either a) cannibalize their business or b) be incrementally not significant revenue c) would cause too much distraction from the main main thing
- Acquire a proprietary data set (Peridodic Labs) in enough volume in a domain of usefulness to eclipse frontier quality. Often happens in robotics, biology, chemistry.
If you do overcome the challenge of building a model that is useful and well priced beyond big labs models, given the huge price of compute, you still need to play in an area where the revenue / compute ratio is signficant and market demand is large enough to payback your compute spend.
It is a difficult game.
1
17
2,248
RT @Indian_Bronson: Concise read on:
Why @clavicular is right about Looks-Maxxing. I think if started a company akin to, or a sub-brand b…
2
70
The less differentiated the product, the hotter the sales rep. This is why the Corgi Girls exist: insurance is a commodity. Good heuristic for where the world is going and Clav understood this.
Inside the thirst-trapping world of ‘Corgi girls’ — unhinged dating demands, bizarre posts about the ‘N-word’ and outrageous online behavior trib.al/60kE8ov
32
29
867
110,850
If at the limit everything becomes about the right to access and the right to actuate on valuable economic opportunities, attractiveness confers a big advantage: I am quite literally 5x more likely to respond to and enumerate my economic problems to a pleasant Stacy all else equal. I am also much more likely to become a customer, because I'd rather have quarterly catch up calls with the pleasant Stacy. Therefore, by being pleasant and attractive, the pleasant Stacy grants the company she works for differentiated access to my demand and right to actuate on it downstream.
This is a thielian "very important truth that very few would like to publicly admit". This ability / class (~access generating human capital) will be massively repriced.
Has anyone scraped company Linkedin profile photos, asked model to classify attractiveness level, and correlated that with some measure of product differentiation? I know scraping Linkedin is notoriously hard, but request for research.
4
2
47
5,706
Netflix (13) + Warner Brothers (30) produced 43 shows on this list!
If the acquisition went through, Netflix would have had a monopsony on differentiated content production.
They knew that if you’re a talent in Hollywood, you either sell to the highest bidder (Netflix) or to the most prestigious studio (HBO).
This would have also been a masterful counterposition against slop, which would have played nicely in their favor, particularly with parents/high value demos.
The New York Times ha finalizado su lista de las mejores series del siglo.
Breaking Bad tomó el primer puesto.
3
1,038
A muse inspires, stokes / enumerates desire. Nominative determinism.
It's not just that Google has too many different surfaces, fiefdoms, too much bureaucracy, etc. It's also the case that most consumers (myself included) don't know what they want / how they'd want to use a personal agent, and Meta is the only large platform that has been natively architected for this.
Since the beginning, they've built products that billions of people are happy to mindlessly interact with. All other technology products require mindful interactions: you have to tell Google what you want, whereas Meta knows what you want!
Meta architected their products such that your mindless interaction with them elicits the behavioral context / data distributions necessary to understand your preferences, desires, needs, etc. They've been on a never-ending quest to build products that, through interaction, generate a universal ledger of consumer needs / desires.
By virtue of having this behaviorial graph / real-time ledger of consumer problems, they're also best positioned to create a proactively useful personal agent: one that can sense problems and route solutions in real-time.
This proactivity is absolutely necessary because most people are bad managers / delegators / don't know what they want! Previously, Meta routed content and ads (mostly for consumer goods) to fulfill consumer desire, now they will route end-to-end solutions against a wider / deeper spectrum of needs and desires.
Today, this including routing requests to real people who can solve your problems end-to-end (e.g, if you want to cancel your phone plan they route your request to a contractor who calls xfinity on your behalf). This of course also produces very useful trajectory / training data that turns human labor today into model labor tomorrow.
The only thing hampering them is that they don't have native access to the platform layer (iOS, Android, GSuite, M365, etc.) But! Muse makes Meta Glasses much more useful, and in turn, Meta glasses make Muse much more useful. So Muse adoption could lead to Meta Glasses adoption and vice versa. If I started to see this dynamic play out, I'd get very excited (both as a customer and investor
1
3
1,882
This is actually pretty smart by OpenAI as a strategy to convince normies to buy the new device: camera quality is very important to influencers!
7
641
In a world of super-capable AI, the price of economic mutation / execution goes to zero. What will matter most is building infrastructure capable of continuously identifying and selecting mutations that confer durable survival advantages.
That means understanding the nature / shape of demand, because demand is the market’s selection mechanism.
Aaru is building this infrastructure, and I’m very bullish on the approach / team.
Simulation has the potential to be transformative, but only if it's accurate. Today, we're sharing results of a comprehensive evaluation across 2,993 questions, with an industry-leading mean TVD of 7.62% and an MAE of 3.53%.
More below, including the full post on our site.
1
39
4,796
This has turned out to be a nice trade so far: NASDAQ up 4%, BSP down 15%. Follow for more alpha.
I also have a galaxy brained long thesis on bending spoons that I will publish soon. Any guesses?
Agents indexing / monitoring / auto cancelling subscriptions actually a pretty massive bear case for Bending Spoons now that I think about it.
1
13
1,527
It's not just that Google has too many different surfaces, fiefdoms, too much bureaucracy, etc. It's also the case that most consumers (myself included) don't know what they want / how they'd want to use a personal agent, and Meta is the only large platform that has been natively architected for this.
Since the beginning, they've built products that billions of people are happy to mindlessly interact with. All other technology products require mindful interactions: you have to tell Google what you want, whereas Meta knows what you want!
Meta architected their products such that your mindless interaction with them elicits the behavioral context / data distributions necessary to understand your preferences, desires, needs, etc. They've been on a never-ending quest to build products that, through interaction, generate a universal ledger of consumer needs / desires.
By virtue of having this behaviorial graph / real-time ledger of consumer problems, they're also best positioned to create a proactively useful personal agent: one that can sense problems and route solutions in real-time.
This proactivity is absolutely necessary because most people are bad managers / delegators / don't know what they want! Previously, Meta routed content and ads (mostly for consumer goods) to fulfill consumer desire, now they will route end-to-end solutions against a wider / deeper spectrum of needs and desires.
Today, this including routing requests to real people who can solve your problems end-to-end (e.g, if you want to cancel your phone plan they route your request to a contractor who calls xfinity on your behalf). This of course also produces very useful trajectory / training data that turns human labor today into model labor tomorrow.
The only thing hampering them is that they don't have native access to the platform layer (iOS, Android, GSuite, M365, etc.) But! Muse makes Meta Glasses much more useful, and in turn, Meta glasses make Muse much more useful. So Muse adoption could lead to Meta Glasses adoption and vice versa. If I started to see this dynamic play out, I'd get very excited (both as a customer and investor
at this point it’s unclear to me whether google can actually execute on the personal agent product it desperately needs as well as meta has or even openai.
this category requires a very specific product dna which meta spent an enormous amount of money recruiting & acquiring.
you need to collapse absurd technical complexity into something simple, friendly, opinionated, & intuitive then tell a compelling story. this is a non trivial task.
also google’s other problem is that its product surface is so fragmented that i don’t even understand how the various orgs negotiate ownership, distribution, etc. i.e does thus live in gmail or gemini or both? ironically having just the google connector makes the problem much easier for everyone else.
i genuinely can’t remember the last time google shipped a consumer product built from the ground up this complex that felt completely coherent end to end. anyway there should be a four alarm code red inside the company esp given how quickly meta can turn muse into a household name.
1
1
19
3,425
