Peter Eibich

Andrew: I'm worried I'll be alone on my birthday

Thames Valley Police: Relax

The Apprentice's Sorcerer Or I how I learned to start worrying and love Claude Code

🤖NEW POST ON AI🤖

Perhaps you're thinking like Brenda from Bristol, 'not another one!' but... I felt that lots of writing about Claude Code doesn't really show you how it works. So I thought well, why not? And along the way some bad literary allusions. 1/n

benansell.substack.com/p/the-appren...

The Apprentice's Sorcerer Or I how I learned to start worrying and love Claude Code

This is a really interesting and helpful piece about working with AI by @benansell.bsky.social
open.substack.com/pub/benansel...

Looking forward to welcoming you to Paris!

We are very happy to support the new edition of this workshop!

We will also give a short presentation on common mistakes when writing economics papers, and we will award a prize for the best paper

We’re very excited to be headed back to the Economics of Mental Health Workshop this June—this year as sponsors! Stay tuned 👀

some of you are looksmaxxing when you need to be booksmaxxing

My impression is that almost everyone (incl. Big Pharma) tends to think of medical research largely happening in the pharma industry, and I think that most conspiracies would not hold up if people would actually meet some of the people doing academic medical research

And again, competitive markets rescue the day!

When I was 12 I thought Phil Collins was the most boring musician ever. When I was 16 I thought he was at least a decent jazz & prog drummer. When I was 26 I thought he was a pretty smart songmaker. Now at 34 I am convinced he might be one of the biggest musical geniuses of the 20th century

I'm not saying this in a critical way, so much as a "my God am I a coal miner who doesn't understand new tech" way, but also what are these agents actually doing?

Measles is one of those nightmare diseases that, if it hadn't been killing people for hundreds of years, people would ABSOLUTELY be pushing conspiracy theories about right now. It's too nasty in too many different directions to seem natural. But oh, boy, it sure is.

This is why we need heterogeneous treatment effect estimation

Conscious decoupling

I was a bit concerned tbh when I saw your posts that you are about to wade into the APC debate

I have been ranting about the APC problem as an entirely imaginary issue to anyone who would listen fot a few years now (surprisingly few people, most just ask "who are you? What are you doing in my house? Are you aware that it's 3 a.m.?").

So I am glad to learn that you are backing me up here

Zumal die Lebenserwartung mit den Arbeitsverhältnissen zusammenhängt. Wer einen gut bezahlten und sicheren Job hat, der lebt länger. Wer wenig verdient, prekär arbeitet und auch noch gesundheitsschädlichen Arbeitsbedingungen ausgesetzt ist, lebt kürzer.

Die Lebenserwartung ist eine soziale Frage.

Happy birthday to Charles Darwin,
patron saint of tired scientists, grumpy fieldworkers, and hating your own manuscript.

On the publication bias discourse, I regret that metascience has become a source of decontextualized, low-res, bean-counting-focused `science is in crisis' narratives. It is largely uncurious abt science, desperately lacking in theory & measurement. I'll quote a few takes I liked & add my thoughts🧵

Difference in time use between working parents of small children vs working non parents. Areas above zero are activities parents do more of; areas below are what they give up. Compiled by analyzing the Census Bureau’s American Time Use Survey. Analysis and chart by Aziz Sunderji, https://homeeconomics.substack.com/p/where-do-parents-find-the-time

Where do parents find the time to parent? Less sleep, work and screens.

Amazing chart feat. in @alphaville.ft.com Further Reading.

homeeconomics.substack.com/p/where-do-p...

What do babies need teeth for anyway? My commiserations...

If you have used Claude Code or similar for research, I'd love to pick your brain on two issues I have been wondering about:
- Data confidentiality: what kind of data do you give it access to?

- Data security: how do you deal with the risk that the agent does something stupid with your data?

Der linke Bereich der Sonne in gelber Farbe. In dieser vergrößerten Darstellung zeigen sich zahlreiche Strukturen auf der Oberfläche und am Rand: Da wechseln hellere und dunklere Gebiete ab, gehen in Wirbeln ineinander über, da strömen hell leuchtende Strahlen teils geradeaus, teils in großen Bögen in die Umgebung, alles sehr dynamisch und spektakulär.

Guten Morgen zusammen - wie gewohnt mit einem aktuellen Bild der Sonne, heute mal wieder per Satellit SDO und kräftig reingezoomt.

My conclusion is that much of the worry about sub-replacement fertility is overstated. Quantitatively, the net effect of even a large fertility reduction on the US economy would be a relatively small decline in the standard of living. Comparing demographic steady states and focusing on the most easily quantified channels, a version of the United States with a total fertility rate of one child per woman would have consumption per capita that was 8.7 percent lower than a version of the country where the TFR was two. In the first four decades of the transition following a decline in fertility below the replacement rate, consumption is actually higher than it would have been if fertility had remained constant. Indeed, much of the sturm und drang regarding the economic effects of current population aging is related to the ending of such a transitory period of good times that resulted from fertility declining from its Baby-Boom highs to near replacement, starting in the 1960s. Finally, it is important to note that any attempt to fix the economic problems stemming from low fertility by raising the birth rate will entail a period of higher overall dependency in the decades that it takes the resulting children to become productive adults.

Low birth rates have modest long run negative effects, after good effects for a few decades, and raising birthrates exacerbates dependency. A giant, apparently permanent, disconnect between expert opinion (e.g., below) versus manufactured panic hype on this issue pubs.aeaweb.org/doi/pdfplus/...

Yes, this is one of the use cases I can clearly see - quality checks for code. Def important and will probably improve research quality

Headline from The Hill reading "How to raise birth rates is the wrong question: Here’s what we should be asking"

One of the most frustrating things about the current US discourse over low birth rates is its laser focus on "how can we raise birth rates?" In this op-ed for @thehill.com, I explain why that's the wrong question and what we should be asking instead. 1/6
thehill.com/opinion/heal...

Thinking about my own work, I am fairly certain that given an AI agent access would be a breach of most of my data access agreements (if not in letter then in spirit). Similar concerns apply to other areas (e.g., using an AI to expand my notes into a reviewers report)

If you have used Claude Code or similar for research, I'd love to pick your brain on two issues I have been wondering about:
- Data confidentiality: what kind of data do you give it access to?

- Data security: how do you deal with the risk that the agent does something stupid with your data?

As someone who has not used it, where is it most helpful for you?

I understand it is great at writing code, but writing analysis code is a small proportion of my time, and using it for data cleaning seems problematic if the data are confidential?