Manhattan, NY
Joined December 2024
- Tweets263
- Following1.5K
- Followers951
- Likes16K
PE firms should build internal RL teams if they want to improve DPI and reduce equity checks. They can i) compound data internally, turn each company into one large eval and use it to expand margins, drive topline, etc... or ii) build RL tasks and envs and license them to labs to drive cash generation if a PortCo is in a tough spot or they need to return cash to juice up IRR (which is a much better solution than a dividend recap imo)
When you have 20% of your portfolio in NVDA
An alert:
@JensenHuang is human. And he is an incredible human. Even incredible humans make mistakes.
He takes full responsibility as he should.
We should all support him and his family.
Onward.
CrossCurrents retweeted
Excited to share Memorilla! Incredible to see @adibvafa @Radii2323 @kvignesh1420 carry this through
A new way of looking at memory, as a separate modality
I worked on initial experiments with RL: can we learn how to better use past experiences by interacting with the world? 🧵
🦍 Introducing Memorilla: memory as a modality for LLMs
Memorilla turns everything a frozen LLM needs to remember into just 16 memory embeddings
⚡ It uses up to 4,212× fewer tokens and is 27× faster than full context inference while more accurate
🎯 Pushes the Pareto frontier over RAG and agentic memory in personalization, long-document and knowledge QA
A thread 🧵:
Couldn’t be more bullish on @terraindustries. @JTLonsdale @kokoxs nailed it.
Insecurity is a massively overlooked bottleneck to financing infrastructure and upstream projects across the Global South.
Africa has some of the richest critical mineral deposits on earth, yet Chinese capital can often underwrite projects that Western institutions struggle to finance under stricter risk mandates.
So if the US is serious about securing critical mineral supply chains, reducing physical security risk is a critical first step.
CrossCurrents retweeted
pre the new website yuvmalik.com
This was most definitely an investment to access alt data for Thiel macro. Im thinking of at least 5 commodities they can trade with this data
Recentemente, o bilionário Peter Thiel investiu em uma startup que coloca coleiras inteligentes em vacas.
Parece piada, mas a empresa já vale mais de US$ 2 bilhões.
Funciona assim: o pecuarista desenha uma cerca virtual pelo celular e o GPS da coleira mantém cada animal dentro da área definida. Quando a vaca se aproxima do limite, recebe um sinal sonoro e uma leve vibração indicando para onde deve voltar. Em poucos dias, ela aprende.
Na prática, o pecuarista consegue mover o rebanho inteiro arrastando uma linha na tela. Sem porteiras, quilômetros de arame ou a necessidade de percorrer o campo.
Mas a parte mais valiosa não é a cerca.
Enquanto o gado pasta, o sistema registra cinco leituras por segundo de cada animal, acompanhando movimento, comportamento e condição física.
Com esses dados, a IA consegue identificar sinais de doenças, período fértil e até prever partos antes que isso seja percebido por alguém no campo.
O nome da IA é Cowgorithm. Ela foi treinada com mais de 7 bilhões de horas de comportamento real de vacas.
Para um leigo, o modelo de negócios parece ser apenas as coleiras. Mas, na realidade, são os dados.
Hoje, mais de 1 milhão de animais usam a tecnologia, por um custo de cerca de US$ 5 a US$ 8 por vaca por mês.
O Brasil tem um dos maiores rebanhos bovinos do mundo, com mais de 200 milhões de cabeças, e boa parte do setor ainda opera com métodos tradicionais.
É uma tecnologia com espaço enorme para crescer por aqui!
Really grateful to ChinaTalk, a leading publication on China and emerging technology, for featuring my piece, “NdFeB Magnets: The Material Bottleneck Behind Humanoid Robotics.” Cool to see it featured by a publication I’ve long followed for its work on China, technology, and industrial capacity.
chinatalk.media/archive
Everyone’s starting to catch on to the real bottleneck in scaling critical industries like robotics, EVs and drones. I wrote about it back in February: xitter.cf/CrossCurrents28/status…
CrossCurrents retweeted
personal ai's saturating the internet will lead to an acceleration in the shift to the dark forest economy
when everyone's personal ai is fighting for a Carbone res / Odyssey tickets, the alpha will go back to irl
> members clubs are a good v1 as tiktok / social media ruined public spaces
imagine this dynamic but for large portions of economic activity that was previously mediated by the internet
firmly in the "i know a guy" era
which room are you in?
CrossCurrents retweeted
Que rápido llegó el enriquecimiento cultural a Ceuta.
CrossCurrents retweeted
Introducing Castle
A company, today, is an unintentional wrapper of macro events: an airline is an embedded oil trade, a solar developer is a tax-credit bet, and a data center is a wager on permits clearing before a moratorium lands. These exposures are intangible, lack historical data, and, therefore, remain unprotected on corporate balance sheets.
On the other hand, there are players who actively seek this risk. As quantitative traders, we modeled these events, and we craved more avenues to express our opinions and to provide liquidity.
Castle exists to bridge this gap. We leverage markets and models to cover the risks that insurers won’t, enabling businesses to focus solely on their execution.
Read about one of our recent trades, in collaboration with our friends at Susquehanna and @Kalshi , as a CNBC exclusive linked below.
We were quants, engineers, and hedge fund managers. Now, our Mission is to reprice one of the world’s oldest and largest markets.
Find out more about Castle’s Mission below
CrossCurrents retweeted
the greatest innovation in investing of the past decade has been the emergence and outperformance of crossovers
while people deride the fomo investing / price inelastic nature of their investing, there is a better game afoot
if you view the private book as proprietary information flow to better position the public book, the picture becomes clearer
like yeah - loss ratio is 99%
"(INSERT FUND) is the next Tiger Global"
but the math actually papers out: if one private co becomes a hectacorn which returns the fund and all the proprietary info flows into the public book and nets 30% IRRs, you would take that all day
why do you think the private returns are so bad compared to NASDAQ???
commoditizing your complement comes to investing
CrossCurrents retweeted
a 10 minute application for a $50,000 fellowship. no strings.
a total of $1,000,000 going out, towards making more magic in the world.
presenting the Basis Fellowship.
run by @markkhrapko ,@thewildstevenp , and spencer green
CrossCurrents retweeted
THE BEAUTIFUL GAME ™️
savvy sensor positioning
^^
sensors soaking in signal
^
act
🔁 spin cycle
This is catastrophic for Botswana, need a polymarket for IMF intervention in <5 years.
Diamonds are the biggest anchor of demand for their currency, so a devaluation could destroy them given they import the majority of essential goods (fuel, food, medicine and machinery). Could be a humanitarian crisis.
Expect the collapse in diamond prices to accelerate substantially over the next few years. The natural-diamond market is shrinking in both size and value and circling the drain following the commoditisation of its product by modern technology. The same thing happened to pearls 100 years ago, but it will be much worse for diamonds: they will become a signal of bad taste rather than wealth, just as pearls went from high society to a matronly cliché within 50 years.
I have been telling friends in the diamond industry to find a new line of work for ten years. It was clearly only a matter of time before lab-grown diamonds became a mass-market commodity, sold at some small premium over production costs—which would themselves continue to plummet.
Since they are identical to natural diamonds, and buyers actually have multiple reasons to choose them instead—no “blood diamonds” and supposedly greater environmental friendliness—most buyers will not care.
Attempts to differentiate the products using $20,000 machines will not help: they are indistinguishable in real life.
There was always going to be some multiple that people would be willing to pay for “natural” over “lab,” but it did not really matter whether that multiple was 2× or 10×. Once lab-grown diamonds can be made for next to nothing, even ten times next to nothing is still very little. People will pay more—perhaps much more—for an identical product because of its backstory, but not infinitely more.
The reason natural-diamond prices have held up reasonably well over the past four years—“only” falling by around 50% in real terms—has more to do with supply cutbacks. Production has been reduced by about 20% over that period, from 120 million to 98 million carats, in a desperate attempt to support prices as the natural and lab-grown markets diverge. If production returned to its previous level, prices would fall even faster.
But it is going to get much, much worse. A trip to Miami will explain why: people who look like gangsters walk around wearing enormous tennis bracelets that would have cost a million dollars 20 years ago but can now be purchased for the price of a used Rolex.
Diamonds are going to become a sign of tackiness—as, frankly, they always should have been—rather than class or exclusivity. And then the game will be up for everyone in the diamond industry.
