Sondre Ulvund Solstad

@clement_molin (To put it perhaps a bit more simply: the claim made does not apply to this chart or to the article.)

@clement_molin Hi, in part for the reason you outline, we--including in this chart--do not count Russian-infiltrated areas as Russia-controlled. Instead, we show (as the chart suggests) the change in Russia-controlled Ukraine.

These circulating figures do not reflect reality. The ISW is creating a very "progressive" map that favors Russia (infiltrations are counted as Russian control). This is how imaginary counter-offensives are invented, portraying Ukrainian advances into territory that has never been under Russian control.

@littmath I find them to be very unpredictable too--almost feels like I'm talking to different people, if that makes sense--extremely careful and creative in some settings, making elementary mistakes the next.

But I do worry that there’s a gold rush mentality and not everything is being checked carefully.

Still underrated how uneven frontier models are within math, IMO. I’ve recently been reading through some of the more interesting solutions to Erdős problems and quite enjoying them—here the models are reliably executing nontrivial ideas, combining known techniques, etc. But…

Reported intercepted attacks, and assumptions about the number of interceptors used per missile, and share of drones shot down by them, make it possible to estimate the total number used: economist.com/international/…

@clary_co Thanks!

Good example of what you can do with publicly available data and some algebra. The Economist examining whether Gulf missile defenses are running low. "Stocks are under severe pressure." economist.com/international/…

And some differences in strikes between week one and week two:

Latest on the war--clear dip in incoming drones to Bahrain, Qatar and UAE. economist.com/middle-east-an…

@LykinsGerry Thermal data from satellites which cannot say why something is very hot only that it is. Explained at link.

???? “High temperature events”. ????

Data has obviously since come in. I process updates 3x daily. So far, it is pointing in the same direction as trends in the charts above. We show a selection of the latest here: economist.com/interactive/20…

I analysed data on 788 distinct attacks, 1,860 missile and drone launches, and abnormal “high-temperature events”, such as fires, at 208 strategic sites. It suggested the war has entered a new phase. economist.com/interactive/mi…

@BenDoBrown @bilawalsidhu when I spoke to them*, I meant to write.

@BenDoBrown @bilawalsidhu Or actually, I was aware of the Sudan Witness Project, but when I spoke it was still quite manual in the end

I built a free tool that shows how night-time lights have changed anywhere on Earth. Here's how it works and why it's useful for monitoring conflict, disasters, development and growth. 👇

@bilawalsidhu @BenDoBrown FIRMS is widely used to detect conflict events. However, it does require a lot of processing/modelling if you want to do it at scale and in an automated way. AFAIK I'm the only one doing the latter.

@BenDoBrown helpful, thank you! most of the false positives i was getting were industrial plants

It is rumoured that Russia's leader will assess whether to attempt another offensive in March. No large city has fallen since Mariupol in May 2022. The towns that have, such as Avdiivka and Bakhmut, were reduced to rubble long before the Russian flag was raised.

The latest casualty estimates—from official and independent sources, as well as The Economist’s own modelling—suggest that Russia is losing far more troops than Ukraine, while gaining almost no ground. economist.com/graphic-detail…

Our cover this week. As I write in the accompanying briefing: where the press is stifled today, corruption should be expected tomorrow. economist.com/briefing/2026/…

@cblatts Ever more vs code here. Some extension usually resolve whatever difficulty I encounter

For Claude coding, any thoughts on VS Code vs Cursor? I like the cursor layout in autocomplete, but I don't think I need a subscription when I don't use the chat and its built-in access to models. I tested it out for a month, but I'm looking for another alternative that lets me run Claude in terminal, see that it might mark down documents, and see my file explorer all in the same screen.

Every November, @TheEconomist ponders the year to come. For this year’s edition of The World Ahead, I wrote about the conflicts to watch in 2026. I highlighted seven—of which one has already taken a major turn. economist.com/interactive/th…