‼️Retail leverage frenzy is also spreading in Taiwan: Margin loans in Taiwan have surged to a record NT$600 billion ($19 billion), more than DOUBLING over the past year. This exceeds the peak set during the Dot-Com Bubble in 2000. At the same time, borrowings backed by stocks and ETFs have hit a separate record, with 16 billion shares pledged as collateral, a figure that has surged nearly 4 times since 2022. The surge in margin debt over the last 12 months dwarfs even the +50% increase recorded in the final year of the Dot-Com Bubble, and exceeds the +94% rise seen recently in South Korea. For many Taiwanese investors, this is their first time borrowing money to amplify their market bets. If the AI buildout slows, the unwind of this leverage will be massive. #Nikhil_BNB #HYPEFalls17%FromRecordHigh @Nikhil_BNB
BREAKING: 🇺🇸🇮🇷 The US has reportedly issued a 60-day general license allowing the sale, production, and export of Iranian oil. With exports estimated at around 2 million barrels per day and oil near $75 per barrel, Iran could generate roughly $9–10 billion in revenue over the period. This could mark a major boost for Iran’s economy and a significant return to global energy markets $BTC #Nikhil_BNB #NakamotoShiftsToBitcoinFocusedBusiness
🔴The entire South Korean and Taiwanese stock market rally rests on just 3 companies: Samsung Electronics and SK Hynix together reflect ~55% of South Korea's Kospi index. Both companies recently crossed trillion-dollar valuations and dominate the production of memory chips used in AI computations and data storage. Meanwhile, TSMC alone accounts for ~42% of Taiwan's Taiex index. TSMC is the world's 7th most valuable company, with a market capitalization of over $2.2 trillion, larger than Tesla or Meta. TSMC also accounted for more than 90% of revenue in the most advanced chipmaking segment last year, according to Counterpoint Research. To put this into perspective, the Magnificent 7 collectively account for only ~33% of the S&P 500. This is INSANE. $SPCXB #Nikhil_BNB @Nikhil_BNB
I keep coming back to something that feels slightly unresolved about health data. Wearables already measure sleep cycles, heart rate variability, movement, and dozens of other signals while we sleep. At the same time, AI is becoming increasingly capable of interpreting those patterns. Yet the more I look at it, the less the challenge seems to be accuracy alone. Most discussions focus on whether an AI interpretation is correct. Better models, larger datasets, and more refined predictions usually become the center of attention. But what feels interesting is that the origin of those interpretations often remains invisible. That’s where ideas like Dream Auditing started making more sense to me. Not because of the analysis itself, but because of the possibility that an interpretation could carry proof of where it came from. In systems like OpenGradient, an output could potentially be accompanied by cryptographic evidence showing which model produced it and whether it remained unchanged. The more I think about it, the less verification feels like a technical feature and more like a trust system. Sleep and cognitive data are deeply personal, and once AI begins interpreting them, the ownership of those interpretations becomes less obvious. I might be overthinking this, but maybe the next challenge for AI is not producing another answer. Maybe it is preserving the history of how that answer came to exist. And if that becomes important, the real question may not be whether we trust AI, but whether we eventually expect every meaningful output to prove itself. $OPG #Nikhil_BNB @Nikhil_BNB