Publications (Google Scholar): scholar.google.com/citations…
For speaking/collaboration, feel free to DM.
B. Branisa | Economics & Data Science retweeted
En 1990, #Bolivia y Perú tenían un PIB por habitante similar. Hoy Perú nos aventaja considerablemente.
Superar la crisis actual es urgente, pero ¿qué Bolivia queremos dentro de otros 35 años?
Mi artículo de hoy en Correo del Sur:
correodelsur.com/economia/20…
In Chile, returning ancestral land to Mapuche communities led them to replace commercial pine and eucalyptus plantations with grasslands for livestock, but it brought no measurable gains in carbon storage or biodiversity. voxdev.org/topic/energy-envi…
B. Branisa | Economics & Data Science retweeted
You want an underrated channel on both mathematics and physics?
You want to learn about Index theory, Riemann surfaces, analytical number theory or the mathematical requirements for QFT?
Then, you should check out this channel by KhalKhali (who i believe is good friend with the one and only, Alain Connes!)
B. Branisa | Economics & Data Science retweeted
“Causal Effects Versus Causal Mechanisms: Two Traditions With Different Requirements and Contributions Towards Causal Understanding”
statmodeling.stat.columbia.e…
B. Branisa | Economics & Data Science retweeted
Mi nuevo post... ¿Dejó Homo erectus una huella genética en los denisovanos y, a través de ellos, en algunos humanos actuales?
¿Hemos encontrado por fin quiénes son los superarcaicos detectados en nuestra historia evolutiva?
B. Branisa | Economics & Data Science retweeted
Read the introduction: sebastiangaliani.com/wp-cont…
B. Branisa | Economics & Data Science retweeted
1/2 We just received the proofs of AI at Your Side: The Student’s Guide to Smarter Learning, forthcoming from Oxford University Press!
Why does this book matter? Because how we learn to use AI may be just as important as the technology itself.
En 1990, #Bolivia y Perú tenían un PIB por habitante similar. Hoy Perú nos aventaja considerablemente.
Superar la crisis actual es urgente, pero ¿qué Bolivia queremos dentro de otros 35 años?
Mi artículo de hoy en Correo del Sur:
correodelsur.com/economia/20…
B. Branisa | Economics & Data Science retweeted
The biggest thing happening in the world right now is physical AI.
i went to bed and asked claude to paint an autumn tree overnight.
at around 4am, it realised that it didn't have some shades -- so it stopped and titrated its own paint: diluting red, sienna and madder step by step on a second palette, then measuring each pocket from a photo.
woke up to this beautiful watercolor painting. 303 brush marks, took about 4.5 hours 🍂 (still working on the speed)
B. Branisa | Economics & Data Science retweeted
This is really worth emphasizing: you can get incredible increases in welfare but the GDP growth numbers would barely move. Economics of how automation translates into aggregate statistics can be non obvious and counterintuitive.
Google DeepMind's @alexolegimas explains how millions of AI agents could transform the economy while GDP growth stays at just 4-5%.
The reason: AI could make production so cheap that the economic boom barely shows up in GDP.
"Let's say you have tons of autonomous organizations, millions of agents. There’s a state of the world where you’re still in the four to five growth number."
"Agriculture got automated. What happens to agriculture as a part of the economy? It went from like 40% to 2%, even though there was way, way, way more agriculture."
"Whatever gets automated becomes cheaper, and if those firms are producing a lot of output for not a lot of cost, the price of that output drops, and hence they might not necessarily lead to huge growth numbers."
"We could see incredible flourishing, incredible amounts of cheap output that we’re all consuming and we’re all happy with, and when you look at the GDP numbers, they’re in the regular sort of ballpark."
@GoogleDeepMind
B. Branisa | Economics & Data Science retweeted
Many years ago, while researching a paper on U.S. monetary policy, I read the testimony before Congress of several extremely prominent economists during the 1960s and 1970s.
I was somewhat saddened to discover that they did not follow standard stabilization principles. When the economy was weak, they recommended expansionary fiscal and monetary policy (the right advice in that scenario). But when the economy was booming, they never supported contractionary fiscal and monetary policy (also the right advice in that situation) because they did not want to “kill the expansion.”
No, sorry. If you want to manage aggregate demand (and I believe we should), you need to understand that this is a balancing act and, even more importantly, that fiscal policy has to be sustainable. Keynes himself was extremely clear about this. Gösta Rehn (1913–1996) and Rudolf Meidner (1914–2005) were even clearer: the government should run a surplus on average to have enough fiscal space to respond to large shocks (wars, financial crises, etc.). It is precisely because we care about those most vulnerable in society that you want to run a surplus. The cries against “austerity” are nearly always a mix of demagoguery and irresponsibility.
I suspect that many of those who defend Keynes most ardently in the public debate are the ones who least appreciate how good an economist he was.
I went to see last night “The Standard of Living”, the play now in London about John Maynard Keynes’s life, and I really enjoyed it. It captures well his life and the times. However I disagree with an important part of the economics, a confusion that is behind the entire economic policy debate in Britain and the rest of Europe today: the confusion between Beveridge and Keynes.
The report by William Beveridge (an LSE economist and ex-director) led to an enormous expansion of the welfare state. That is often confused (also in the play) with Keynes’s views, as if the large and continuously growing post-war welfare state had anything to do with demand stabilization.
Keynesian demand management is what Andrés Velasco (also an LSE prof), as Finance Minister in Chile, did during the 2000s copper boom: he ran surpluses of 7% of GDP and more in 2006 and 2007 and put the money in a stabilization fund abroad. This was 12% of GDP by the end of 2008, when public debt was 4% of GDP. Public sector workers burned his effigy in Santiago asking for it to be spent. When the copper price collapsed along with the global economy, this surplus allowed him to introduce a massive stimulus package worth 2.8% of GDP, larger relative to the economy than the American one, which allowed him to stabilize Chile’s economy through the financial crisis. “Being a Keynesian means being one in both parts of the cycle,” he told The Economist. Today, very few governments are doing this. It is “whatever it takes” in crisis, and “welfare is untouchable” when the crisis finishes. This leads to a continuous ratchet up of spending that strangles any productive activity.
Keynes was in favour of using debt to stimulate demand, but under no circumnstances to finance permanent welfare spending. He wanted an ordinary budget for the state and then a capital budget that would be used for the demand management and that would be invested only in productive capacity. While Beveridge was about social redistribution, an issue of social policy, Keynes was about stabilization. The play and a lot of commentary seem to conflate both things and use demand management to justify permanent structural social policy.
I think this is pretty instructive for what’s happening in the UK and in the rest of Europe today. There is a widespread confusion that it’s neoliberalism which is hollowing out the state, when in fact the UK’s public spending burden is at around 45.6% of GDP this year, and the OBR expects it to stay around 45% for ten years running, the longest sustained period of spending this high since the Second World War; EU governments spend 49.5% of GDP (Eurostat, 2025). Outside wartime and the pandemic, the state has never been this large for this long. Neither neoliberalism nor austerity is at fault for the lack of investment.
Governments cannot hit the fiscal rules without touching the triple lock, and other welfare entitlements (free transport for the over 60s here in London!) that most MPs consider untouchable. What gets cut all the time to make the numbers add up is infrastructure, energy, grids, R&D, innovation, etc. The UK’s gross fixed capital formation is at 18-19% of GDP, which is the lowest in the G7, and the general government investment is also behind the OECD average.
The debate misunderstands the risk of pension sustainability. It is not that at some point in the future pensions become unsustainable. It is that today, their relentless growth eliminates the space to invest in infrastructure, innovation and the rest. Sadly, no party dares to make this argument. And even economists who defend the entrepreneurial or innovative state cannot bring themselves to state the obvious: that there is a huge opportunity cost in having one of every 5 new cars sold in the UK being paid by the state as a disability benefit, or in having free transport for all over 60 year olds.
B. Branisa | Economics & Data Science retweeted
🚀 New {tidyterra} release on CRAN!
✨ Improvements & bug fixes
🤖 Experimental AI chatbot on pkgdown
Try the chatbot, ask away & share your feedback!
#rstats #rspatial
dieghernan.github.io/tidyter…
B. Branisa | Economics & Data Science retweeted
Con base en este tweet de @karpathy , le pedí a Claude que armara una página para mirar papers que me pueden interesar. Lo alimenté Inicialmente con los tweets que he publicado a lo largo de estos años y con la idea de Karpathy.
Me muestra 12 papers por página, cada uno con su título, la revista y por supuesto un enlace. Les asigna un color y un ícono según el tema y una frase que resume por qué podría interesarme.
Puedo marcar cada paper como leído, tuiteado o "no me interesa". Esas marcas le enseñan qué priorizar en la próxima actualización, que por ahora haré a pedido, pero que se podría automatizar.
Cómo lo hizo: identificó los cientos de papers que cité en mis tweets y de qué revistas venían. Bajó todo lo que esas ~50 revistas publicaron en el último año, unos 22.800 artículos. Los comparó con lo que tuiteé, quedaron 337, y de ésos eligió 93.
No escribí una línea de código: solo escribí un prompt y luego tuve un breve ida y vuelta. La primera impresión es que es bastante cercano a lo que necesitaba. Con el uso, iré mejorando mi propio algoritmo.
La idea es que reemplace lo que hago hasta ahora: leer los índices de cada número de muchas revistas, lo cual es tedioso y arroja un paper interesante cada decena o centena de títulos que leo.
Acá va una captura de pantalla y en el próximo tweet el prompt.
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
hackertube.org/
- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
B. Branisa | Economics & Data Science retweeted
Feats of human excellence:
- Wish You Were Here (album)
- 3Blue1Brown's linear algebra series
- TSMC
- Googling Strangers and Kentucky Bluegrass
- Mr. Tambourine Man
- The Moon landing
- Messi's 91-goal season
- Lost season 3 finale
- Siddhartha by Herman Hesse
- Animals (album)
- The founding of the US
- Zlatan's 41m bicycle kick
- Runaway by Kanye
- General relativity
- A Song of Ice and Fire
- Sultans of Swing (live at Hammersmith)
- LLMs
- Cold fact (album)
B. Branisa | Economics & Data Science retweeted
bro forgot to mention @DavidDeutschOxf deriving probabilistic expectations from non-probabilistic axioms... literal magic (arxiv.org/abs/quant-ph/99060…)
Feats of human excellence:
- Wish You Were Here (album)
- 3Blue1Brown's linear algebra series
- TSMC
- Googling Strangers and Kentucky Bluegrass
- Mr. Tambourine Man
- The Moon landing
- Messi's 91-goal season
- Lost season 3 finale
- Siddhartha by Herman Hesse
- Animals (album)
- The founding of the US
- Zlatan's 41m bicycle kick
- Runaway by Kanye
- General relativity
- A Song of Ice and Fire
- Sultans of Swing (live at Hammersmith)
- LLMs
- Cold fact (album)
🤖🤔😱
MIT and Harvard built an AI simulation containing 8.3 billion virtual people, roughly the entire population of planet.
They call it MatrAIx.
It’s a population-scale simulation running on frontier models, GPT and Claude among them. The longer description took me a while to believe.
Start with the dataset, Persona 8B. 8.3 billion synthetic profiles, each one specified across 1,290 dimensions. Background, psychology, spending habits, how comfortable they are with technology, what they do on a Tuesday night, the small behavioral quirks that make one 34-year-old nurse in NYC different from another.
A spreadsheet of 8 billion rows would be a curiosity. What they did next is the part that matters.
They turned the profiles into agents and dropped them into four environments: surveys, AI chat interfaces, live web browsing, and native desktop and mobile apps. Then they ran real products through the crowd. New features, software updates, pricing changes.
And they logged everything. How long an agent paused after seeing a price go up. The exact point at which it gave up on a broken checkout. How much lag it would sit through before closing the app.
The number I keep coming back to is 91.5%. That’s how often the agents stayed in character, behaving the way their assigned psychology and demographics said they should. Not perfect. But for billions of agents across four environments, closer than I expected.
Here’s what that does to how products get built.
Right now, if you want to know whether a feature will land, you ship it to a slice of users and wait. Or you pay a research firm and wait longer. Either way the answer arrives in weeks, from a sample you hope is representative.
With something like this, you run the feature past 8 billion simulated reactions overnight. One server. Results by morning, broken down by every one of those 1,290 dimensions if you want them.
I have questions about how well a synthetic crowd predicts a real one. 91.5% is persona adherence, not forecast accuracy, and those are different things. But the direction is clear, and it’s coming faster than most product teams are ready for.
B. Branisa | Economics & Data Science retweeted
Even if one might not agree with everything in it, it’s a must-read paper!
pubs.aeaweb.org/doi/pdfplus/…
B. Branisa | Economics & Data Science retweeted
'Monte Carlo is an extremely bad method; it should only be used when all alternative methods are worse.' -- Alan Sokal
B. Branisa | Economics & Data Science retweeted
Much to learn (for me) from this paper.
Mind the Ladder is now on TMLR:
jmlr.org/tmlr/papers/
Abstract:
openreview.net/forum?id=01WK…
If you are into JEPA and CAUSALITY a must read, I am doing more work of course @yudapearl and @ylecun thanks for the patience!



































