iiinova — where your digital self keeps evolving

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

This project was made possible by an awesome team, including @dholtz.bsky.social, Alex Martin Richmond, Christopher Ong, Sonny Tambe, and Ronnie Chatterji This is our first time sharing this paper publicly, so we'd love to hear your thoughts as we polish it.

010

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

We have a bunch more results in the paper. If you're interested in a looking at a higher-level overview, as well as some nice visualizations of our findings, we also have an accompanying blog post: Paper: cdn.openai.com/pdf/5d1e1489... Blog post: openai.com/index/how-ag...

cdn.openai.com

100

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

As a result, measures of output are up substantially--the median researcher at OpenAI now produces 56x as many output tokens across ChatGPT and Codex as they did in November, and the median employee on the legal team now produces 13x as many.

100

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

We also see that workflows are evolving--about one in ten users now runs three or more concurrent agents at some point each week, a share that has grown rapidly over time.

100

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

People are using agentic tools in increasingly complicated ways: in May, one in four individual account users made a request that would have taken a skilled human more than 8 hours to complete. This was extremely uncommon as recently as December.

100

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

Although most people still use traditional chatbot interfaces (think ChatGPT), a growing number of people are switching to agentic interfaces, which can perform complex work on your behalf (think Codex). Codex accounts for a much greater share of output tokens than of users.

100

Drew Johnston @drew-johnston.bsky.social · 25/06/2026

How are people using agentic AI tools to change the way they work? My team at OpenAI looked into this, using data from Codex that lets us measure agentic AI usage among individual users, organizational account users, and workers at OpenAI.

100

Reposted by Drew Johnston

Andreas Bjerre-Nielsen @andbjn.bsky.social · 29/07/2025

1/ What does the social fabric of an entire country look like? We built a nation-scale social network of Denmark — 7.2 million people, 1.4 billion ties, 14 years of data. Here’s what we found 👇 📄 doi.org/10.1038/s415... #NetworkScience #Sociology

doi.org

Unveiling the social fabric through a temporal, nation-scale social network and its characteristics - Scientific Reports

Scientific Reports - Unveiling the social fabric through a temporal, nation-scale social network and its characteristics

12513

Drew Johnston @drew-johnston.bsky.social · 21/07/2025

I'm excited to present at @ic2s2.bsky.social in Norrköping! Wednesday at 10, I'll give a lightning talk about how to measure cross-class social connections (almost) everywhere on Earth. I'll also have two posters up in the Weds session. If you're here, I'd love to chat social networks, etc!

150

Drew Johnston @drew-johnston.bsky.social · 20/06/2025

If you want to learn more, our paper describing the research is available here: drew-johnston.com/files/cross_... (with bonus maps here: drew-johnston.com/files/cross_...) and all the data is available to download at HDX: data.humdata.org/dataset/cros...

drew-johnston.com

011

Drew Johnston @drew-johnston.bsky.social · 20/06/2025

You can read the full article here: economist.com/internationa... Many thanks to the team behind the research ( Mike Bailey, @ayushkumar.bsky.social, Theresa Kuchler, and Johannes Stroebel) and the team at the Economist (@ainsliejstone.bsky.social ) for making this all possible!

economist.com

Can men and women be just friends?

The answer matters more than you think

131

Drew Johnston @drew-johnston.bsky.social · 20/06/2025

On the flip side, they find that labor force participation gaps are a better predictor of the rate of more marginal (top 200) friendships in a place.

120

Drew Johnston @drew-johnston.bsky.social · 20/06/2025

The Economist published an article today about my team's research on measuring social ties between men and women. The article had cool original analyses, including showing that the rate of cross-gender ties among close (top 5) friends is predicted by an index of sexism (a 🧵)

183

Reposted by Drew Johnston

Drew Johnston @drew-johnston.bsky.social · 30/05/2025

My team's paper on cross-gender friendships is out today in the AEA Papers & Proceedings. We use data from Facebook to show where connections between men and women are more (blue) and less (red) common, in almost every country, and release the data publicly! A 🧵 of maps:

132

Drew Johnston @drew-johnston.bsky.social · 30/05/2025

We'd love to see you use the data in your own research. You can download it at a variety of granularities here: data.humdata.org/dataset/cros... and find the paper here: doi.org/10.1257/pand... or on my website: drew-johnston.com/files/cross_...

020

Drew Johnston @drew-johnston.bsky.social · 30/05/2025

Our paper describes the construction of the data, but we think there are still a ton of stories to tell from people who have more local context than we do. For instance, I'd like to know more about the differences we see here!

111

Drew Johnston @drew-johnston.bsky.social · 30/05/2025

132

Reposted by Drew Johnston

Owen Winter @owenwntr.bsky.social · 24/03/2025

I had a sneak peek at some data which is being published by Meta today, showing that Britain is less divided by class than you might expect: www.economist.com/britain/2025...

economist.com

New data show that the class divide in Britain may not be so wide

They make the country look better than America

119226

Drew Johnston @drew-johnston.bsky.social · 17/03/2025

We use self-reported gender from individuals' profiles.

000

Drew Johnston @drew-johnston.bsky.social · 17/03/2025

Would love to see something like this, but it would take a bunch of data work to pull off. Perhaps it's possible in the future if there's enough interest!

020

Drew Johnston @drew-johnston.bsky.social · 17/03/2025

In the data release, we provide information about how common cross-gender ties are among the closest friendships relative to among less-close friendships; it might be possible to use this to say something about cross-country patterns in within-family friendships relative to more distant ones.

010

Drew Johnston @drew-johnston.bsky.social · 17/03/2025

Great questions! I think the first few "closest" friendships are likely to be partners or family, though this is just an educated guess based on the fact that the closest few friendships display weaker gender homophily in most countries (see Figure 1 in the paper).

110

Drew Johnston @drew-johnston.bsky.social · 14/02/2025

Realized I forgot a to include a map of cross-gender social ties in North America--details on the methodology (and a link to the data) can be found in the original thread!

050

Drew Johnston @drew-johnston.bsky.social · 13/02/2025

My post is for the top 200 friends. And yes, it's very likely due to the different number over which the statistics are calculated! Check out the appendix to the paper, which has many maps plotted separately by N top friends: drew-johnston.com/files/cross_...

110

Drew Johnston @drew-johnston.bsky.social · 13/02/2025

Thank you so much for your interest! Glad to hear that people find this interesting--one small note though, 1 here actually indicates no bias, not 0.5! I've copied the formula for the values here. You can see more details in the paper: drew-johnston.com/files/cross_...

200

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

This project sprung out of a collaboration with Mike Bailey, Theresa Kuchler, @ayushkumar.bsky.social, and Johannes Stroebel and would not have been possible without them!

150

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

We think our data is an appealing way to measure attitudes on gender, particularly in countries not often surveyed. If you're interested in working with this data, download it here: data.humdata.org/dataset/cros... The paper (out soon in AEA P+P) is available here: drew-johnston.com/files/cross_...

1238

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

Across countries, the Cross-Gender Friending Ratio is strongly predictive of gender differences in labor force participation. Within countries, we also find a strong correlations with gender attitudes in the World Values Survey, such as opinions about women's suitability for political office.

172

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

We measure gender differences using the Cross-Gender Friending Ratio, the ratio of female friends in men's networks to the share of female friends in women's networks in a given place. Men almost always have a lower share of female friends than women do, but the degree varies across countries:

151

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

Have you ever wondered how social networks differ by gender? Check out my team's new dataset, which uses Facebook data to measure regional differences in social networks by gender all across the world! A 🧵 with examples, a description of our methodology, and a link to download the data:

25514

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

000

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

000

Drew Johnston @drew-johnston.bsky.social · 12/02/2025

100

Drew Johnston @drew-johnston.bsky.social · 14/01/2025

If you're considering an econ predoc, Theresa Kuchler + Johannes Stroebel are hiring a predoc to work on social networks projects, starting fall 2025. They are wonderful to work with and do very cool research (see www.nature.com/articles/s41...) . Apply at: apply.interfolio.com/161883

nature.com

Social capital I: measurement and associations with economic mobility - Nature

Analyses of data on 21 billion friendships from Facebook in the United States reveal associations between social capital and economic mobility.

030

Drew Johnston @drew-johnston.bsky.social · 03/01/2025

I'm presenting "Social Capital Around the World" Saturday @ 8am at AEAs. We use data from 2.5 billion Facebook accounts to measure cross-class and cross-gender friendships globally, and explore how they connect to downstream outcomes like intergenerational mobility. www.aeaweb.org/conference/2...

aeaweb.org

American Economic Association

020

Drew Johnston @drew-johnston.bsky.social · 11/12/2024

Would love to be added! Going to be rolling out a new paper next month :)

120

Drew Johnston @drew-johnston.bsky.social · 02/12/2024

Awesome paper, congratulations! Martin Koenen and I just came out with a paper that might be of interest to you. We use FB microdata to get a broader but shallower look at the diffusion of shocks (mostly focusing on migration). Would love to chat some time! drew-johnston.com/files/Social...

drew-johnston.com

020

Drew Johnston @drew-johnston.bsky.social · 21/11/2024

Would love to join the list if you still have space!

110

Drew Johnston @drew-johnston.bsky.social · 18/11/2024

Hey! I'd love to join this, hoping to share some new work on here soon :)

110

Drew Johnston @drew-johnston.bsky.social · 15/11/2024

Hey there! Any chance I could be added if you still have space? Working on a few things I'm hoping to promote in the new year :)

100

Drew Johnston @drew-johnston.bsky.social · 23/10/2024

Would love to be added! hoping to promote some work on here soon :)

110