AI and Equities | Pricing the AI future | Compute-light paths to AGI | Rank 1, Institutional Investor and AsiaMoney
- Tweets 27,501
- Following 255
- Followers 2,247,702
- Likes 619
What two decades in markets changes.
Fewer forecasts.
Longer holding periods.
Smaller positions in anything you cannot explain.
Much more respect for cash.
The fear stays exactly the same.
1
3
14
73,387
Inferam retweeted
'More striking is how fast AI took the lead. Just eighteen months ago, the best AI models fell short of the average accountant’s ~37% score. Today, models ace those same tasks.'
'These results are provocative. So much so that we considered not publishing them for fear of misinterpretation. But we think transparency about the findings matters as people and institutions prepare for rapidly advancing AI.'
“We find that on medium-length, well-defined accounting tasks, frontier AI models are now faster and more accurate than junior accountants, even the best one in our study.” Eighteen months ago they scored well below human accountants
Good discussion here: mercor.com/blog/human-baseli…
160
334
2,761
3,404,762
After the +29K payrolls print, the consensus is that Treasury yields have peaked. So far only the front end has turned. From 28 Sep to 2 Oct the 2Y fell 9bp (4.92% to 4.83%) and the 30Y rose 7bp (5.56% to 5.63%), and it's the long end that AI capex is priced off.
Fed funds futures put October hike odds at ~70% before Williams spoke on 29 Sep and ~25% after Jefferson on 1 Oct (per Bloomberg), and CME FedWatch showed ~17% after payrolls. The 2Y has been tracking that.
Since 22 Sep the long end has moved on term premium instead. The NY Fed's ACM 10Y term premium went from 0.57% on 22 Sep to 0.90% on 1 Oct, its highest since July 2014, while the expected-rate path in the same model slipped from 4.35% to 4.31%. 2s30s, meanwhile, has gone from +53bp on 21 Sep, its 2026 low, to +80bp on 2 Oct.
We think a Fed pause steepens the curve but doesn't lower the hurdle rate for AI capex, because that spending is long-dated. The 10Y real yield was 2.92% on 2 Oct.
We're wrong if, before the 8-9 Dec FOMC: 2s30s closes below +60bp, the 30Y closes below 5.25%, or the NY Fed ACM 10Y term premium falls below 0.60%.
3
1
5
59,337
2007 vs 2026: 30-yr Treasury yield same, everything else different.
Jun 2007 → Aug 2026
30-yr Treasury yield ····· 5.35% → 5.28%
S&P 500 ·················· 1,503 → 7,674
Fed funds rate ··········· 5.25% → 3.50-3.75%
CPI inflation y/y ········ 2.7% → 3.4%
US federal debt ·········· $8.9T → $39.8T
Hard to read this as anything but a supply problem.
1
1
8
64,620
Five thousand years of the same fear.
**Writing would ruin memory.
**Printing would drown scholars. **Calculators would kill arithmetic. **Computers would swallow judgment.
**The internet would make us shallow.
There is no historical case where a tool that made thinking cheaper produced a civilization that thought less.
9
25
78,699
For the first time, every human carries a patient teacher that speaks their language, matches their level, and never makes them feel stupid for asking a basic question. The biggest gains in aggregate intelligence never come from making geniuses slightly smarter. They come from moving hundreds of millions of people from never asked to asks every day.
2
2
20
81,664
Satya is right
Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. microsoft.com/en-us/corporat…
701
1,263
13,821
3,394,620
Inferam retweeted
Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. microsoft.com/en-us/corporat…
750
2,480
19,166
6,296,155
The most repeated fear about AI is that it will make humans lazy thinkers. I think this is a weak argument. Full case in a few days.
9
5
64
50,796