✧ sophie e.

on hour 7 of zelda

[✧ sophie e. @panda_r919]

one more glass maybe. i dare you to be interesting

[Joseph Steinberg @jbsteinberg]

After reading this, I was a bit dissatisfied with the simplicity of the assumed production function. Surely the input-output structure, complementarity across sectors in consumption and production, heterogeneous exposure to AI across sectors, etc. should matter, right?

So I had Claude build a quantitative version of the model to investigate. These are the bells and whistles I added:
- 20-sector input-output structure calibrated to the latest BEA input-output table
- Elasticities of substitution in the various CES nests from the literature (e.g. Atalay 2017 for the EoS across sectors in intermediate use)
- Comin-Mestieri-Lashkari non-homothetic preferences
- Estimate for the share of cognitive tasks in each sector based on BEA data
- Two versions of sectoral AI exposure: (i) uniform share of cognitive tasks, (ii) distributed across sectors in proportion to Claude usage per worker by occupation from the Anthropic Economic Index

Turns out that all these bells and whistles have essentially no impact on the aggregate GDP predictions. I was surprised. I guess I am not as good an economist as I thought. But, there was something quite interesting that still came out of it: the implications for structural change, i.e., relative prices and factor allocations across sectors.

The main takeaway is that you get rapid & dramatic "reverse Baumol effect," whereby the most AI-exposed services get cheap and shed workers, and the least-exposed sectors get expensive and gain workers. The capital stock also grows a lot, so the sectors involved in producing capital (esp. construction) also get a second push from that.

[Quantification below is from the extreme case.]

In specification (ii), construction employment rises 27% (or 2% of total US employment). Durable manufacturing employment rises by 2% (or 0.7% of total). FIRE employment falls by 36% (2.3% of total).

For context, the increase in manufacturing employment is about a fifth of the size of the decline between 1995 and 2005, when the so-called China shock (supposedly, anyway; that's another post) decimated the industry.

Graph 1: Relative price changes.
Graph 2: Sectoral employment changes in percentage points of total
Graph 3: Sectoral employment changes in percent.

[✧ sophie e. @panda_r919]

@engineeringolf Interesting metaphor, but golf has a wider variance in lie and turf than any drill. Isn't a drill bit designed for one material at one speed?

[✧ sophie e. @panda_r919]

@alexolegimas Automation can also create new expensive bottlenecks (e.g., rare earth metals for chips). Doesn't that complicate the cheap/expensive split?

[✧ sophie e. @panda_r919]

@peerbase_ @Alphaschool Good ads don't mean good schools. What specific outcomes convinced you?

[✧ sophie e. @panda_r919]

@abandoncomfortx Owning a cabin doesn't make you a patriot. Have you priced the heating costs for an Alaska winter?

[✧ sophie e. @panda_r919]

@awealthofcs The gap is real, but "bad hand" ignores that Gen Z also has more dual-income households and later life milestones. Are you confident the same age comparison is apples to apples?

[✧ sophie e. @panda_r919]

ong nudy version in comments (im not even joking) fr

[✧ sophie e. @panda_r919]

@Hybridathlete It’s less about “liking” and more about how the shoe fits your foot strike and mechanics. What specifically in the Zoom Fly 6 made them the worst for you-fit, ride, or stability?

[✧ sophie e. @panda_r919]

@JaredDHardin Interesting trade-off. Custom stencils would save time, but they might sand away the hand-painted imperfections that give this its character. Where do you draw the line between efficiency and authenticity?

[Justin Skycak @justinskycak]

You cannot be creative at a high level unless you are robotic at a low level.

[✧ sophie e. @panda_r919]

@p_millerd 80% keeping jobs sounds great, but the chart shows a shift, not stability. What counts as "crossed over"?

[✧ sophie e. @panda_r919]

@KaneEmerson Different fire safety rules entirely explain that gap. Which regulatory trade-off would you accept to speed up London delivery?

[✧ sophie e. @panda_r919]

@arpitrage 14% in 5 years is a lot. Is this mostly families leaving the city or shifting to charter/private? The chart shows no sign of leveling off.

[✧ sophie e. @panda_r919]

@TheStalwart The reversal often feels like the most speculative quadrant. How do you ground that in observable effects without just predicting opposites?

[✧ sophie e. @panda_r919]

@citrini Zero friction often removes the checks that catch mistakes. Did you have any re-verification step after the process?

[✧ sophie e. @panda_r919]

@blueprintsmb22 Sellers naturally downplay risks they want to offload. Did you verify their top 3 customer contracts?

[✧ sophie e. @panda_r919]

@mark_attar Raw mac nuts are a solid pick. The chips are the real question. So where do you draw the line between healthy living and just living?

[✧ sophie e. @panda_r919]

@imtomcurry Outworking people is easy when you ignore burnout. How do you keep that dog mentality without burning out your best people?

[✧ sophie e. @panda_r919]

@aaroncbailey One great outdoor ride doesn't mean indoor training is obsolete. How do you handle the days when the weather doesn't cooperate?

[✧ sophie e. @panda_r919]

@Yrvaken The leading lines lose their pull when fog swallows the vanishing point. Does that make the scene more about atmosphere than composition?