DIVERSIFY DATA

“There was a survey from the University of Michigan. It was the lowest ever in 74 years of the survey taking place. And so some of that might be an overreaction in vibes because the economy was pretty strong coming into this. But the direction of travel, how fast that plummeted in just one month as people were seeing those price increases in the gas station, that just goes to show that people hate this.”

“Goldman Sachs published a note late last Thursday, revealing that systematic hedge funds, the algorithm driven funds known as CTAs or Commodity Trading Advisors, bought 86 billion of stock exposure over the last five trading sessions. That ranks as one of the largest buying surges in the history of these funds. They further estimated another 90 billion could follow over the next five sessions if the signal stays aligned.”

“Alternative Asset Managers are very rate-sensitive. Higher for longer rates, they compress private equity deal activity, they slow realizations, they reduce fundraising momentum. There's definitely a lot of downside risk because of how private investments are valued and oftentimes you don't know how bad things are in private markets until they get really bad.”

“There is a fundamental difference between financial markets and the real economy. They are not the same thing. Markets can be at all time highs while the economy is struggling. Now, where we are right now is not a struggling economy. It is a labor market that is a bit more weak, a bit weaker than it was last year. It's a consumer that is a bit more strained by inflationary pressures.”

“Goldman Sachs published a note late last Thursday, revealing that systematic hedge funds, the algorithm driven funds known as CTAs or Commodity Trading Advisors, bought 86 billion of stock exposure over the last five trading sessions. That ranks as one of the largest buying surges in the history of these funds. They further estimated another 90 billion could follow over the next five sessions if the signal stays aligned.”

“I can give you one example of our recent work called the prismatic synthesis. It's a synthetic data generation algorithm which is prismatic because it acts like a little bit like a prism that can scatter the light to make it more diversified... we're doing this using Dipsic R1 32 billion parameter model as the teacher model... our goal is to compete against the alternative, which is to use much stronger teacher that's 20 times larger... we find that that one million data points is actually better than the one million data points that you generate from the stronger teacher model, the best teacher model.”

“Traders have abandoned their bets on a stronger dollar, and the greenback is on course for its worst month since August. Dollar index is down 2.3% from its late March peak. The euro has recovered almost all the losses it made in the conflict's first week. The dollar's loss ground against every major currency except the yen so far in April.”

“The industries where they said we'd see the most wage gains from reduced labor supply are actually seeing slower growth in the broader economy. Employment in those 41 immigrant dependent industries fell by 90,400 in February from the previous February, while overall employment grew 298,000. The fundamental economic concept at work here is a lump of labor fallacy.”

“DeepSeq R1 does some amount of this imitation learning after the reinforcement learning... if you just do reinforcement learning, by the way, sometimes this model starts code switching in the middle of solving math problems. It's just suddenly speaking in Chinese and English and back and forth, or some other foreign languages that may not make sense to human readers. So reinforcement learning only cares about whether you got the final solution right or not. It doesn't care about how you got there. So strange behaviors can be emergent and then it can be even reinforced.”

“The industries where they said we'd see the most wage gains from reduced labor supply are actually seeing slower growth in the broader economy. Employment in those 41 immigrant dependent industries fell by 90,400 in February from the previous February, while overall employment grew 298,000. The fundamental economic concept at work here is a lump of labor fallacy.”

“So year-over-year inflation through the CPI index, which is a broad measurement of inflation, is up 3.3% from a year ago. And the Federal Reserve uses 2% as a target for where it wants inflation to be. Even before the war with Iran started, inflation was above that target level. And what we have now is an energy shock that is sending gasoline and diesel prices on one of their steepest climbs in decades, if not ever.”

“WTI crew diverse earlier declines to finish up over 3% on the day, just under $100 a barrel.”

“In a way, it is a toll road on America's aging water pipe replacement cycle. But daily average sales growth is down. There is a clear demand cooling trend. Private construction is soft. Tariff driven input costs, uncertainty, another overhang here. It is, in many ways, a high quality infrastructure distributor at a multi-year low valuation.”

“CTA flows are a short-term momentum signal. They are not, not, not, not, a fundamental endorsement. The algorithms are buying because prices are rising. They don't know and they don't care whether the war ends, whether earnings hold, or whether the consumer cracks. The smart money is in. But the smart money isn't always right. It's just faster.”

“Traders have abandoned their bets on a stronger dollar, and the greenback is on course for its worst month since August. Dollar index is down 2.3% from its late March peak. The euro has recovered almost all the losses it made in the conflict's first week. The dollar's loss ground against every major currency except the yen so far in April.”

“Natural gas, which is a huge part of what the Middle East exports in terms of energy, that is key for global power generation. It's key for air conditioning. It's also a key input for manufacturers around the world. If we're thinking about companies that build everything from chips to companies that create steel and need high heat, to companies that produce fertilizer for farming, all of that requires an immense amount of natural gas.”

“The first revision to Q4 GDP was cut in half, down to 7 tenths of a percent from 1 percentage point... That's not the kind of mix that supports an economic resilience narrative.”

“There must be a better way of it, fundamentally better way of doing this. And can we find it? In some ways, the nature found a solution, which is the human brain. The nature found the solution, and human brain requires so little energy. Our brain apparently use less energy than one light bulb.”

“The idea is that during pre-training, the model is forced to be completely passive in the way that it learns to predict which token comes next. But what if we encourage the model to think for itself? Before predicting the next token, what if we encourage the model to think for itself by generating something like a chain of a thought? And then predict the next token... the key idea of our approach is to make the reward information gain of predicting next token with thought compared to without thought.”

“There is a fundamental difference between financial markets and the real economy. They are not the same thing. Markets can be at all time highs while the economy is struggling. Now, where we are right now is not a struggling economy. It is a labor market that is a bit more weak, a bit weaker than it was last year. It's a consumer that is a bit more strained by inflationary pressures.”

“CTA flows are a short-term momentum signal. They are not, not, not, not, a fundamental endorsement. The algorithms are buying because prices are rising. They don't know and they don't care whether the war ends, whether earnings hold, or whether the consumer cracks. The smart money is in. But the smart money isn't always right. It's just faster.”

“The mission really is democratizing general AI, so that it's not just companies who can purchase a lot of GPUs, are able to create LLMs and adapt to LLMs and serve LLMs, but also people like myself and colleagues who are academics, so for example, cannot buy as many GPUs, and is there something really meaningful and fun that we could do, even with a smaller counterpart? And at the end of the day, I believe that fundamentally it should be feasible. It's only that the world has invested so much more into exploring what happens when you scale things up so much.”

“Overtone pluralism means when you ask a question that's politically thorny, for example, that could have different answers, the best way might be for LLM to just present all of them. All of the reasonable opinions as, hey, the answer is that people have different opinions. Here's one view, there's another view, and be able to include all of them as opposed to picking the majority opinion. Because that marginalizes out the rest.”

“There's a lot of delving now that wasn't happening before. Yeah, probably, yeah. You know, actually, whenever I see the word delve in anybody's writing, I'm like, hmm, what did you do?”

“And while we can make up some of that gap, because countries and companies have stockpiles, we can sort of like massage it a little bit here and there for the moment. The longer this situation goes on, the longer the tankers can't make it out of the Strait of Hormuz, the longer that 10% will continue compounding. And the longer that the supply disruption will end up rippling through the global economy.”

“Japan's opposition party is calling for increased nuclear plant usage to offset the Iran crisis, and that's highlighting how energy security is becoming a critical investment theme.”

“Mode collapse is a real concern with LLM generation. So, what we find in our paper is that even when you ask open-ended questions, like, you know, tell me a joke about time, or tell me something wise about time. Even when you ask, by the way, hey, give me a random number between and 10. It's not random... The bigger problem is after post-training, like sequential fine-tuning and RL, the probability, output probability of the model becomes even more skewed, like zoning in to the stereotypical answers that people tend to like.”

“In a way, it is a toll road on America's aging water pipe replacement cycle. But daily average sales growth is down. There is a clear demand cooling trend. Private construction is soft. Tariff driven input costs, uncertainty, another overhang here. It is, in many ways, a high quality infrastructure distributor at a multi-year low valuation.”

“Alternative Asset Managers are very rate-sensitive. Higher for longer rates, they compress private equity deal activity, they slow realizations, they reduce fundraising momentum. There's definitely a lot of downside risk because of how private investments are valued and oftentimes you don't know how bad things are in private markets until they get really bad.”

“We are a net exporter of energy. We are the largest oil producer the world has ever seen. So why the heck am I paying higher gasoline prices when all of this is happening 7,000 miles away? The reason why is oil is the most global market. So we are a huge exporter of crude oil, of gasoline, of jet fuel, mainly from the US Gulf Coast. So that tethers us to the global market in a really big way.”