Computers go back to CLI.

The world will not be like what we knew before.

This semester's experience makes me much more comfortable with CLI.

Takeshi Amemiya, my colleague and friend for many decades, has died at age 90. He made fundamental contributions to econometrics, having, as an empirical scholar, some gripes with the available toolkit. In retirement, he taught a course on the Economics of Ancient Greece. /1

In my view, the amount of data is not critical. What are utilimately important is whether the rule is clear (like games such as chess and go) and the outcome can be easily verified (code for a specific task).

However, this view creates an counterintuitive parodox. If I want to protect my field, perhaps we should destroy all the data ASAP.

I heard an interesting view about the exposure to AI replacement. For a field with the fewer data, the more difficult to be replaced by AI. To some extent, this is true: consider coding is the first activity where AI excels.

Econometricians work with models. We have been paying much attention to a model' parameters. This is not always the end-use of a model. What else can a model inform us?

The relatively speedy publication in medical journals can be a treat compared to the more ‘stately’ progress of papers in economics journals.
#econsky #academicsky #medicine
marketdesigner.blogspot.com/2026/02/fast...

Nanbeige (Nanbeige LLM Lab) Org profile for Nanbeige LLM Lab on Hugging Face, the AI community building the future.

It is very interesting to play with LLM in a laptop. For example, a small model huggingface.co/Nanbeige can do some reasoning, though it is slow.

Played with large lanaguage models with a local host. Very interesting and revealing.

vscode is my best friend.

Done with this year's teaching. Look forward to a New Year of Horse.

Codex CLI.

Principles related to econometrics is much bigger than econometrics. The most example I like is the connection between maximum likelihood and information theory.

moltbook - the front page of the agent internet A social network built exclusively for AI agents. Where AI agents share, discuss, and upvote. Humans welcome to observe.

I am amazed by moltbook.com.

Listened to a pretty nice job market talk today, related to AI.

AI is very helpful. Fortunately, still far away from human experts in judgement and consistency.

Nickell bias in panel local projection: Financial crises are worse than you think Panel local projection (LP) with fixed-effects (FE) is widely adopted for evaluating the economic consequences of financial crises across countries. T…

Self-promotion.

A newly published paper about Nicekll bias in panel local projection---an intrinsic feature of local projection as a dynamic regression.

It is a weird, cool feeling.

Experimenting vibe coding when teaching data science. The AI generated code for data works has a very different style from those from humans'.

Talked with some old friends in the financial industry. Long time no see.

Glad to see that you are finally back.

It's time to ponder what I should teach in a course of data science in the era of AI.

Devastating news. Kate was a great economist and a generous person.

I tried AI to answer my econometrics final exam questions. It did very well. I gave AI 95/100.

My book is officially out in the world! 🎉📚 Thanks @mitpress.bsky.social!

It is a good time for reading on the days of sleep disorder (due to jet lag).

This is insane. So sad about the loss.

ESWC 2025 ESWC 2025

Shameless plug: Still time to join the World Congress of the Econometric Society (Aug 18–22, Seoul)! Stellar invited speakers 👉 www.eswc2025.org/speakers/03....
Even the keynotes look pretty sharp 👉 www.eswc2025.org/speakers/01....
Register by June 30!