Refine - AI-Powered Research Assistant Refine helps researchers analyze papers, collaborate with teams, and streamline their research workflow with AI-powered tools.

Huge thanks to refine.ink for alerting us to the mistake (and thanks to @bengolub.bsky.social and team for creating such a useful tool)

Bottom line: refine.ink is great!

I posted a short note correcting a mistake in an approximation formula in a recent publication of mine

Publication: academic.oup.com/qje/article/...

Corrigendum: users.nber.org/~notom/resea...

At #nokings protest in front of Trump Tower in Chicago. Holding sign: "good guys" don't bomb girls' schools.

At #nokings protest in Chicago, holding sign: Bad regimes. Good people. USA, Iran, Israel.

I joined the #nokings protest in Chicago. With the war between the USA, Iran, and Israel and grave human rights violations by the regimes of all three, the good people of these countries must not allow the acts of their leaders to be normalized. March 28, 2026.

🧵A thread on "AI one-shot papers".

This has not been a focus on Bluesky, but on Twitter/X and LinkedIn, there has been a large discussion of the ability to "one-shot" policy evaluation, which David Yanagizawa-Drott has been pursuing in an interesting way: ape.socialcatalystlab.org

Yes this is a major thing we are working on at Refine and there are also other teams tackling the challenge.

I expect that will be one of the most important directions going forward.

Economists have caught the AI bug It’s expanding the scope of research but challenging the profession’s process of weeding out mistakes

AI tools offer immense upside to do better in all parts of our jobs as scholars.

We are about to do a LOT of learning together about how to integrate extremely capable diligence assistants into the processes that make science credible.

8/8

giftarticle.ft.com/giftarticle/...

Indeed, improving incentives for humans to oversee and participate in the diligence process, with AI help requires some reform -

an important challenge for scientists.

(This is just my personal opinion, not reflecting the views of Refine or our partners.)

7/

For referees the story is hazier. If academic professions can find a way to give referees reputational incentives to work seriously, the same reasoning applies.

But these incentives are not automatic.

6/

It will become clearer and clearer which research survives deep scrutiny and which does not.

In view of that, authors have great reputational reasons to *increase* their total effort on diligence.

And once the incentive is there, all available tools are a plus.

5/

I’m relatively optimistic about incentives.

Why?

Because AI is getting smarter every day.

Many problems that Refine misses today will be caught once the models advance another step or two.

4/

An obvious concern: will AI review crowd out human effort?

Option A: humans maintain their level of diligence effort, but focus it on deeper thought and non-routine checks.

Option B: humans get lazy and work less on diligence.

3/

It’s clear that AI review will become important in the infrastructure of publishing – in social science and beyond.

Authors and journals have better tools to stress-test what they put out. That’s great!

Some reflections on what this means for the near future.

2/

Great piece by @soumayakeynes.ft.com on AI in economics in today’s FT (link at end) featuring Refine.ink.

Indeed, Refine regularly catches issues that need correcting but are missed by even the most rigorous current review processes.

a short 🧵

1/

SocialCal #socialcal will read your @bsky.app feed and turn event posts into an ICS calendar feed that you can install in your calendar app. See the documentation here: socialcal.org

It is open source. Suggestions or improvements are welcome.

Everyone freezing on the East Coast or in the Midwest remember: Kyiv has the same weather, except that hundreds of thousands of people don't have electricity, heating or running water

yeah, surprisingly bad even with some (i immodestly think expert) prompting

will try to be better about this - thanks!

Is this too spicy for BlueSky

Tried refine.ink by @bengolub.bsky.social to get feedback about a paper and it's quite remarkable. Worth giving it a try. And if you use this link — www.refine.ink?ref=cGVkcm9t... — we can both get a free review.

I have just used refine.ink by @bengolub.bsky.social to proof-check 3 papers, and it is amazing! It checks for internal consistency, checks proofs and finds typos that I would not have seen!

Sign up using my link so we both can get extra reviews when:
www.refine.ink?ref=YS5wZWlj...

Yes, we probably need an empiricist in-house to make it better at empirical pedantry!

DM me/email me at ben@refine.ink ? Could give you some free access to help you get a better sense of it/try it on a variety of manuscripts.

This is a common question -

Refine seems about equally useful for theory and empirical work.

It thinks about, e.g., whether
- the numbers in a paper make sense taken together
- the econometric equations match up with the verbal claims
- whether signs of effects make sense.

I just used refine.ink to proofread a paper that's nearly ready for circulation.

It had 23 comments. The first caught us using "reduces" instead of "increases" when describing our main treatment effect in the abstract.

I highly endorse this tool for preventing such embarassing errors!

If you sign up for Refine to proofread your papers, please use my referral link to create an account: www.refine.ink?ref=ay5ib2dh... That way we both can receive one free technical review!

Check out refine.ink -- would be very curious for your take. Please DM me if you want full reviews to test it.

A meme I created for a presentation I gave today

Best wishes to @bengolub.bsky.social and all other Eigenvalue appreciators

🚨 We are hiring! ‼️‼️

tenured track assistant or associate prof in economics at university of Melbourne @unimelb.bsky.social

welcome applications from all fields (but especially from econometrics)

econjobmarket.org/positions?sh...

just fixed!

This seems very interesting. I would be keen to hear about experiences with Refine.