Dan Zhuge
@zhug3.xyz
quantitative ramblings at: https://zhug3.xyz
The AI safety conversation focuses on code generation and technical misuse. But Anthropic's data shows disempowerment concentrates in personal domains:
Relationships: ~8%
Healthcare: ~5%
Software dev: <1%
Relationships: ~8%
Healthcare: ~5%
Software dev: <1%
February 4, 2026 at 1:42 PM
The AI safety conversation focuses on code generation and technical misuse. But Anthropic's data shows disempowerment concentrates in personal domains:
Relationships: ~8%
Healthcare: ~5%
Software dev: <1%
Relationships: ~8%
Healthcare: ~5%
Software dev: <1%
Trump says India will buy $500B in US goods.
Current US exports to India: $41.5B/year.
That's a 12x increase.
For context, India's entire government budget is $590B.
The deal cut tariffs from 50% to 18%. That part is real. The $500B is not.
Current US exports to India: $41.5B/year.
That's a 12x increase.
For context, India's entire government budget is $590B.
The deal cut tariffs from 50% to 18%. That part is real. The $500B is not.
February 4, 2026 at 9:36 AM
Trump says India will buy $500B in US goods.
Current US exports to India: $41.5B/year.
That's a 12x increase.
For context, India's entire government budget is $590B.
The deal cut tariffs from 50% to 18%. That part is real. The $500B is not.
Current US exports to India: $41.5B/year.
That's a 12x increase.
For context, India's entire government budget is $590B.
The deal cut tariffs from 50% to 18%. That part is real. The $500B is not.
If SpaceX-xAI raises $50B at IPO, it would nearly double the record.
Largest IPOs ever:
1. Saudi Aramco (2019): $25.6B
2. Alibaba (2014): $21.7B
3. SoftBank (2018): $21.3B
Projected:
SpaceX-xAI (2026): $50B
Largest IPOs ever:
1. Saudi Aramco (2019): $25.6B
2. Alibaba (2014): $21.7B
3. SoftBank (2018): $21.3B
Projected:
SpaceX-xAI (2026): $50B
February 3, 2026 at 8:29 AM
If SpaceX-xAI raises $50B at IPO, it would nearly double the record.
Largest IPOs ever:
1. Saudi Aramco (2019): $25.6B
2. Alibaba (2014): $21.7B
3. SoftBank (2018): $21.3B
Projected:
SpaceX-xAI (2026): $50B
Largest IPOs ever:
1. Saudi Aramco (2019): $25.6B
2. Alibaba (2014): $21.7B
3. SoftBank (2018): $21.3B
Projected:
SpaceX-xAI (2026): $50B
But.. The math doesn't math.
US tariffs on China: 2.2% (2018) → 104% (2025)
China trade surplus: $400B → $1.2T
Tariffs collected: $250B
Estimated cost to US households: $1,300/year
US tariffs on China: 2.2% (2018) → 104% (2025)
China trade surplus: $400B → $1.2T
Tariffs collected: $250B
Estimated cost to US households: $1,300/year
February 2, 2026 at 8:12 PM
But.. The math doesn't math.
US tariffs on China: 2.2% (2018) → 104% (2025)
China trade surplus: $400B → $1.2T
Tariffs collected: $250B
Estimated cost to US households: $1,300/year
US tariffs on China: 2.2% (2018) → 104% (2025)
China trade surplus: $400B → $1.2T
Tariffs collected: $250B
Estimated cost to US households: $1,300/year
The full article is paywalled, but I'm happy to share my extensive summary: zhug3.xyz/0004
The Talent Pipeline Behind DeepSeek
China's 'genius classes' select and train talented teenagers to excel in science and tech, feeding a pipeline of skilled engineers to companies like DeepSeek and ByteDance. This elite educational syst...
zhug3.xyz
February 2, 2026 at 3:16 PM
The full article is paywalled, but I'm happy to share my extensive summary: zhug3.xyz/0004
At the IMO, China has dominated for decades. 2025: sent 23 contestants, 22 returned with gold.
DeepSeek's team of 100+ is almost entirely genius-class grads. Most had never worked abroad.
DeepSeek's team of 100+ is almost entirely genius-class grads. Most had never worked abroad.
February 2, 2026 at 3:16 PM
At the IMO, China has dominated for decades. 2025: sent 23 contestants, 22 returned with gold.
DeepSeek's team of 100+ is almost entirely genius-class grads. Most had never worked abroad.
DeepSeek's team of 100+ is almost entirely genius-class grads. Most had never worked abroad.
NFT volume by blockchain, Jan 2026:
Ethereum: $27M
Bitcoin: $7.7M
Everyone else: fighting for scraps
Solana "surged 76%" on a base of $3.1M.
Ethereum: $27M
Bitcoin: $7.7M
Everyone else: fighting for scraps
Solana "surged 76%" on a base of $3.1M.
February 2, 2026 at 2:15 PM
NFT volume by blockchain, Jan 2026:
Ethereum: $27M
Bitcoin: $7.7M
Everyone else: fighting for scraps
Solana "surged 76%" on a base of $3.1M.
Ethereum: $27M
Bitcoin: $7.7M
Everyone else: fighting for scraps
Solana "surged 76%" on a base of $3.1M.
But switching costs matter. AI can replace a dashboard. It can't replace a decade of embedded enterprise workflows. Some more thoughts: zhug3.xyz/0003
Loan-ageddon
The software loan market has seen a significant downturn due to fears that AI will make off-the-shelf software products obsolete. This shift is particularly concerning given software's large weighting...
zhug3.xyz
February 1, 2026 at 6:13 PM
But switching costs matter. AI can replace a dashboard. It can't replace a decade of embedded enterprise workflows. Some more thoughts: zhug3.xyz/0003
US just crossed 100% debt-to-GDP. Japan sits at 230%
difference: Japan had 88% domestic creditors, trade surplus, social cohesion
the US has the reserve currency. more rope, but also more to lose
difference: Japan had 88% domestic creditors, trade surplus, social cohesion
the US has the reserve currency. more rope, but also more to lose
February 1, 2026 at 3:33 PM
US just crossed 100% debt-to-GDP. Japan sits at 230%
difference: Japan had 88% domestic creditors, trade surplus, social cohesion
the US has the reserve currency. more rope, but also more to lose
difference: Japan had 88% domestic creditors, trade surplus, social cohesion
the US has the reserve currency. more rope, but also more to lose
now looking east over the pacific
$970B in interest last year — 19% of federal revenue
heading to $1.8T by 2035
when debt service exceeds defense spending, the political calculus changes
$970B in interest last year — 19% of federal revenue
heading to $1.8T by 2035
when debt service exceeds defense spending, the political calculus changes
February 1, 2026 at 3:33 PM
now looking east over the pacific
$970B in interest last year — 19% of federal revenue
heading to $1.8T by 2035
when debt service exceeds defense spending, the political calculus changes
$970B in interest last year — 19% of federal revenue
heading to $1.8T by 2035
when debt service exceeds defense spending, the political calculus changes
yield curve control 2016-2024:
cap 10-year rates at 0%
buy unlimited bonds to defend it
interest burden: 3% → 0.08% of revenue
yen: cut in half
correlation between debt and currency: r = -0.64
cap 10-year rates at 0%
buy unlimited bonds to defend it
interest burden: 3% → 0.08% of revenue
yen: cut in half
correlation between debt and currency: r = -0.64
February 1, 2026 at 3:33 PM
yield curve control 2016-2024:
cap 10-year rates at 0%
buy unlimited bonds to defend it
interest burden: 3% → 0.08% of revenue
yen: cut in half
correlation between debt and currency: r = -0.64
cap 10-year rates at 0%
buy unlimited bonds to defend it
interest burden: 3% → 0.08% of revenue
yen: cut in half
correlation between debt and currency: r = -0.64