iiinova — where your digital self keeps evolving

Miklós Koren @koren.mk · 17/02/2026

My favorite graph from our paper: the Cone of Adoption.

Miklós Koren @koren.mk · 17/02/2026

Lots of exciting work by @davidautor.bsky.social @avigoldfarb.bsky.social @lukasfreund.bsky.social @joshgans.bsky.social and many others, capturing the "directional" nature of AI. We think job flexibility is key and best modelled as "collaboration" not "automation."

010

Miklós Koren @koren.mk · 17/02/2026

4. AI adoption depends on both worker and job characteristics. The same worker may use AI in one job but not another. Think academia vs consulting vs tech firm. 5. Adoption is faster when jobs and workers are more flexible, which is the case for knowledge work. Relation to lit:

100

Miklós Koren @koren.mk · 17/02/2026

This simple insight leads to a lot of predictions. 1. Absolute advantage not enough for AI adoption. AI needs a comparative advantage relative to worker's current tasks. 2. No adoption up to some threshold, quick adoption afterwards. 3. Some workers only partially adopt.

110

Miklós Koren @koren.mk · 17/02/2026

Coworkers don't change the production function. They add to the set of outcomes that are feasible. This formulation is unusual in labor, so we borrow Production Possibility Sets from international trade. Worker + AI = convex hull of two PPSs.

100

Miklós Koren @koren.mk · 17/02/2026

GenAI is not a machine and shouldn't be modelled as such. Given the range of tasks it can do if properly supervised, it is more like a coworker. We explore the economics of GenAI adoption among knowledge workers in a new paper: koren.mk/publications...

koren.mk

The Directions of Technical Change

Generative AI is directional: it performs well in some task directions and poorly in others. Knowledge work is directional and endogenous as well: workers can satisfy the same job requirements with di...

143

Miklós Koren @koren.mk · 12/02/2026

As an economist who grew up under socialism, I agree.

130

Miklós Koren @koren.mk · 03/02/2026

Vinyl skills: 1. What is vinyl? 2. It has two sides. 3. Can place the needle right before the bridge of Karma Police.

010

Miklós Koren @koren.mk · 29/01/2026

Indeed, (almost) every regression is a GROUP BY query. Amazing job @gmcd.bsky.social

010

Miklós Koren @koren.mk · 28/01/2026

We can debate the magnitude but I don't see how you can say vibe coding does not weaken user engagement. If Claude (etc) selects and imports a library instead of a human user, how is that user engaged with the developer / user community of the library?

371

Miklós Koren @koren.mk · 26/01/2026

Our paper with @gaborbekes.bsky.social @julianhinz.bsky.social and Aaron Lohmann is/was on top of HN. 15 minutes, I guess.

183

Miklós Koren @koren.mk · 25/01/2026

My favorite German veterinarian jargon: tier spezifisches Eigengeruch I keep repeating it to my dog.

011

Miklós Koren @koren.mk · 25/01/2026

Also, I don’t see the harm the title could do even if picked up by a journalist who didn’t read the paper. Open source has a brittle business model. It grew fast but it can also collapse fast. People should be aware of that. Don’t take the “random person in Nebraska” for granted. xkcd.com/2347/

xkcd.com

Dependency

040

Miklós Koren @koren.mk · 25/01/2026

2. I very much favor @saywhatyoufound.bsky.social titles (doesn’t show in my pubs i know). This is not a speculation. This is not a pun. Not clickbait. This is what we found. Watering it down to “Quality Selection and Entry into Open Source in the Age of AI” would have been a disservice to readers

120

Miklós Koren @koren.mk · 25/01/2026

The Melitz model performs very well in predicting selection into exporting, and how competition induces exit. We need more work on the empirical calibration, but the key assumption theta > sigma holds for sure.

110

Miklós Koren @koren.mk · 25/01/2026

If we want to treat the model seriously, though, it has to be put together seriously. Informed by empirical evidence, institutional details, understanding of existing similar models and their performance. Our model is based on the Melitz model. We realized it captures the relevant mechanisms best

120

Miklós Koren @koren.mk · 25/01/2026

1. Economic theory, especially general equilibrium models, provide useful predictions for changes that are too large and systemic to be studied in a small experiment. Theory is not a way to capture one particular mechanism with math. It is a coherent way of thinking about how the mechs interact.

130

Miklós Koren @koren.mk · 25/01/2026

Good discussion, thank you! You’re raising two points: 1. how much can we learn from an economic model 2. what is our responsibility in giving paper titles I agree these are important and non trivial questions. We have grappled with them both. My views, may not be fully shared by coauthors:

130

Miklós Koren @koren.mk · 24/01/2026

There will be almost no need to actually type syntactically correct source code. But a lot of need to think like a software engineer. But: How do you teach programming without using programming languages? Would be curious to see a vibe coding first curriculum.

130

Miklós Koren @koren.mk · 24/01/2026

Yes, and congestion, too. When a lot of contributions are vibed, it is hard to do a good job on all task.

110

Miklós Koren @koren.mk · 24/01/2026

Fair points. The model has dev cost going down due to AI, the result is that in the race between that and the demand collapse, the latter wins. We also speculate on what alternative business models can help. Vibe Coding Kills Open Source Unless We Make Drastical Changes would be accurate, though.

120

Miklós Koren @koren.mk · 24/01/2026

All audiophiles are rich, but not all rich are audiophiles.

040

Reposted by Miklós Koren

Gábor Békés @gaborbekes.bsky.social · 23/01/2026

Vibe coding kills open source. Our most direct title yet. @koren.mk @julianhi.nz @aaron-lohmann.bsky.social Theory paper with numbers and policy recs. First at arxiv.org/abs/2601.15494 Comments welcome. @ceu-economics.bsky.social @kiel.institute

25825

Miklós Koren @koren.mk · 24/01/2026

It’s not click bait if it’s true. The only condition needed is that ai-assisted usage and direct usage of the *same* software package are closer substitutes than different packages. We are sure this holds (though plan to do more data work to confirm), that’s why we opted for the strong title.

121

Miklós Koren @koren.mk · 19/11/2025

Words you only hear in a vibe coding session: "Now let me deploy and test"

010

Miklós Koren @koren.mk · 15/11/2025

Hungarians will understand what this is for.

010

Miklós Koren @koren.mk · 13/11/2025

It ended being a rave party, but hey, had fun anyway.

000

Miklós Koren @koren.mk · 11/11/2025

My rainbow window sticker approves of the music. Prepping for KGLW concert tomorrow.

152

Miklós Koren @koren.mk · 10/11/2025

Now it just keeps apologizing. I could imagine a thought process of "no large RDBMS needed" but, I mean, it has "SQL" in the name.

010

Miklós Koren @koren.mk · 10/11/2025

I had an AI agent refer to SQLite as "noSQL-like DB." Looks like it was trained on Medium/LinkedIn bullshit but I told it in no uncertain terms that this does not pass with me.

130

Miklós Koren @koren.mk · 17/10/2025

Today I was playing with econometrics (matrices and stuff) as well as relational data modelling (how are different entities related) and realized something. > Econometrics is data modelling without the data. I made a similar point in a talk some years ago (github.com/korenmiklos/...)

github.com

010

Miklós Koren @koren.mk · 16/10/2025

This morning I heard my son hum Nirvana in the bathroom. Not sure if I should be proud or worried.

010

Miklós Koren @koren.mk · 12/10/2025

Claude Sonnet 4.5 is currently working on the script that will get me what I want. Users don't need chatbots. They need to seamlessly integrate AI into their workflows to minimize their pain. If Google doesn't understand this, they are in trouble.

050

Miklós Koren @koren.mk · 12/10/2025

I subscribed to Google Gemini and cancelled it within 5 minutes. I wanted to create a Google Form from a survey instrument in Google Doc. Gemini cannot do that. It suggested to go to Google Forms and create one myself, using the Doc as a guide.

140

Miklós Koren @koren.mk · 01/10/2025

This. Today my coauthors complained that Github does not properly render the math from my @obsidian.md notes. By the time we finished the meeting, Claude Sonnet 4 has built a nifty CLI tool “ghmd” that converts Obsidian markdown to github flavor. Complete with unit tests and pipx installable.

010

Miklós Koren @koren.mk · 29/09/2025

When you start your commit message with bigfux: instead of bugfix: you can be sure it was a major bug.

051

Miklós Koren @koren.mk · 22/09/2025

The way AI helps human-to-human communications is underrated. I used to be picky about collaborating on technical projects. Like, "don't send me an email, send me a pull request." Now I do some of my best work just talking to coauthors. AI transcribes, AI summarizes using the project context.

010

Miklós Koren @koren.mk · 15/09/2025

How to make sure you ran reproduce all steps of your research months later? We demoed our CLI tool bead.zip that saves code and data together at RSECon at Warwick. Slides bead.zip/rsecon25

021

Miklós Koren @koren.mk · 15/09/2025

I have been working with coding AI agents several hours a day in the past couple of months. You get to know their style. Grok is like Elon Musk. Moving fast and breaking things at 10x speed. Not apologizing either.

010

Miklós Koren @koren.mk · 29/07/2025

For the narrow subset of #econsky people clinging to their Stata and crunching big datasets with @duckdb.org, I have vibe coded a DuckDB extension that reads .dta files. koren.dev/software/ Give it a spin and report issues! (Thanks, @anthropic.com Claude Code and the Python Pandas implementation.)

-- Load a Stata file as a table
SELECT * FROM read_stata_dta('path/to/your/file.dta');

-- Count observations
SELECT COUNT(*) FROM read_stata_dta('data/survey.dta');

-- Filter and aggregate
SELECT
region,
AVG(income) as avg_income,
COUNT(*) as obs_count
FROM read_stata_dta('data/households.dta')
WHERE age >= 18
GROUP BY region;

042

Miklós Koren @koren.mk · 25/03/2025

Always excited to hear Elhanan Helpman. If you're in Vienna, come by. events.ceu.edu/2025-05-07/p...

events.ceu.edu

Public Lecture: Elhanan Helpman

072

Miklós Koren @koren.mk · 25/03/2025

Why I love OneDrive, part 78. My local folder is called /Users/koren/OneDrive\ -\ Central\ European\ University\ \(CEU\ GmbH\ Hungarian\ Branch\ Office\)/ Luckily, I can symlink.

151

Miklós Koren @koren.mk · 14/03/2025

Good point, this never occurred to me.

000