Sibo Liu

Associate Professor of Economics @ Hong Kong Polytechnic University
Innovation | Political Economy

[AEA Journals @AEAjournals]

Forthcoming in the AER: "Difference-in-Differences with a Continuous Treatment" by Brantly Callaway, Andrew Goodman-Bacon, and Pedro H. C. Sant'Anna. https://t.co/DK3l1laHyQ

[Jacob Edenhofer 🇪🇺 🇺🇦 @edenhofer_jacob]

Glad to have this @CESifoNetwork working paper out — it’s taken a while, but happy with the first draft. More detailed thread to follow.

[NBER @nberpubs]

About 26 percent of the 2025 US tariff increase passes through to consumer prices, from Mary Amiti, Sebastian Heise, and @deweinstein https://t.co/rw1hwEOLNS

[Thiemo Fetzer 🇪🇺🇺🇦 - same handle elsewhere @fetzert]

Very relevant paper, and one may add that is why perceived (in) security (possibly legacy trauma induced) across different socioeconomic strata is so important. This can create varieties of what I refer to as unholy electoral coalitions between groups across the income distribution….

[Thiemo Fetzer 🇪🇺🇺🇦 - same handle elsewhere @fetzert]

For some related longer pieces:
https://t.co/GWTQ7aDqwk
https://t.co/LNLEavgyFO

[Princeton Bendheim Center for Finance @PrincetonBCF]

@MarkusEconomist is publishing a new edition of his award-winning book "The Resilient Society," offering a fully revised and expanded guide to how resilient thinking helps us face a world of unknowns.

Now available for pre-order: https://t.co/DEKwD90yRb

[Miles Deutscher @milesdeutscher]

I automated my entire life using Claude.

My team of AI agents manages every business SOP, task, and workflow that I used to do manually.

Automate your entire life with Claude (copy me):

[Kevin A. Bryan @Afinetheorem]

Really cool paper on the incidence of higher wage demands in unionized industries. One q, though, that I didn't see in the paper (@annastansbury?) How do we know these are rents and not quasirents from past investment? In the latter case, I worry a bit about this rent division.

[NO1ennn @N01ennn]

MICROSOFT QUIETLY BUILT ONE MEMORY MODULE THAT PLUGS INTO ANY AI AGENT AND CUTS ITS CONTEXT BY UP TO 100X

no retraining. no task-specific redesign. you just plug it in

Microsoft's insight: agents don't need to remember what happened. they need the knowledge pulled from it - the facts and the reusable skills, not the raw logs

so PlugMem stops storing history. it compiles experience into a knowledge graph, then feeds the agent only what matters

the result:
> up to 1-2 orders of magnitude fewer tokens
> highest decision-utility per token of any system tested
> one unchanged module beats task-specific memory on chat QA, multi-hop search, and web agents

more memory was never the answer. better memory is - and Microsoft just shipped it as a plug-in

bookmark this. the agents that win won't remember more, they'll remember smarter.

[Matt Dancho (Business Science) @mdancho84]

Two Anthropic senior AI engineers just made Karpathy's loop 1000x better with "Graph Engineering"

Then they dropped 11 page PDF.

Get it here (free):

[Jey @0xJeyx]

this is f*cking gold

Andrew Ng put the whole loop-to-graph shift on one page.

Loop engineering was the last thing everyone learned. Graph engineering is what it turns into, and the whole thing changes the moment you stop running one loop and wire several into a graph.

I dropped it into my setup. The very first run was different.

Not a little better. A different kind of output.

Claude stopped handing me an answer and started handing me work a second agent had already torn apart.

Bookmark it before your feed swallows it.

Read it now, then check the guide below.

[Mahax @Mahaximus_]

Google Brain founder, Andrew Ng:

"Prompting will die in 6 months.

Loops and Graphs are what's replacing it."

In 2 hours, he shows exactly what the best engineers already build instead, and how to start building it yourself.

The missing piece most people skip: how to connect those loops into a graph that compounds every time it runs.

Watch it, then read the full guide on loops and graphs below.

[Behavior Change for Good Initiative @BehaviorChange]

Can AI predict how people will respond in social science experiments? A new @Nature paper by Team Scientist @RobbWiller and collaborators had LLMs act as survey respondents.

Comparing AI-generated results to real results from 70 studies, the AI-generated treatment effects were strongly correlated with the actual experimental effects. However, the AI systematically overestimated effect sizes.

https://t.co/HR3fLzSMMz

[NBER @nberpubs]

Chinese patenting reflects innovative progress, isn’t concentrated in corporate giants, relies heavily on universities, and isn’t driven by inventors with US experience/training, from Josh Lerner, @nummoose, @DimitrisPapan20, Amit Seru, and Zunda Winston Xu https://t.co/wrgDT2SMfY

[Andrew White 🐦‍⬛ @andrewwhite01]

I've written a new blog post exploring how long scientific claims live, by analyzing the history of 3,444 claims over the last 50 years. You can use this to predict the acceleration in scientific progress - how much faster the turnover of facts is by decade. 1/4

[The Review of Economics and Statistics (REStat) @restatjournal]

Legal IP rights are not important to most U.S. firms but are more important to large firms that invest in R&D. Just Accepted new paper by Filippo Mezzanotti and Timothy Simcoe https://t.co/giotdX9afM

[Kevin A. Bryan @Afinetheorem]

Wednesday morning, so "innovation/entrep/econ of AI research tracker" I mentioned has updated. Scrapes 200+ sources, autofixes source errors, grabs relevant research, rewrites abstracts in standard format, autouploads. ~50 papers/week. A few this morning that caught my eye: 1/10

[Adam C Palmer @ac_palmer]

Must-read article for anyone whose job involves *thinking*

https://t.co/88ezsrjMgz

"The problem with writing with A.I. is that it’s mentally enfeebling...
What, uniquely, does writing do? ... It compels thinking in ways that silent contemplation or spoken language rarely can."

[Claudio Ferraz @claudferraz]

A lot of it ends up being about political economy. Just producing evidence on policies is not enough