Tessera
Folgen

Michal Ovádek

@movadek.bsky.social

Lecturer in European Institutions, Politics and Policy at University College London https://michalovadek.github.io/

495 Posts2612 Follower974 Folgt
Michal Ovádek@movadek.bsky.social · 3w ago

shortly pulling up into Liverpool for the Elections, Public Opinion and Parties (EPOP) conference! First time and excited to be presenting joint work with Umut Yüksel on comparative estimates of trust in courts from over 160 countries. Meanwhile, Stuart Perrett is presenting our experimental paper

BildBild
Uta Staiger@ukstaiger.bsky.social · last mo.

Already looking forward to sharing a panel with @anandmenon.bsky.social AND @jillongovt.bsky.social AND @alexhh.bsky.social AND @movadek.bsky.social to discuss all things EU-UK, where we're now & where we're headed. 8 Oct @uclspp.bsky.social & @uclei.bsky.social. www.ucl.ac.uk/social-histo...

Policy & Practice Seminar: Europe at a Crossroads

The EU in a Changing World Order - what comes next for the bloc in terms of security, cooperation, and defence? And what’s next for UK-EU relations ten years after the Brexit vote?

ucl.ac.uk

Michal Ovádek@movadek.bsky.social · last mo.

as per Eurostat data, bond yields in Hungary dropped by almost 2 pct points (~25%) since Viktor Orban lost power in the April election, even as yields are slightly up over the same period in the rest of the EU

Bild
Vicente Valentim@valentimvicente.bsky.social · 4mo ago

Our department has three calls for tenure track positions. IE is a fast growing group of scholars doing top-notch work, which offers very attractive conditions in a great European city. If you are looking for an academic job in Europe, definitely do consider applying.

IE Political Science Department@iepolisci.bsky.social · 4mo ago

We are hiring! We have 3 tenure track positions open: - Comparative Politics: apply.interfolio.com/188149 - International Relations: apply.interfolio.com/188150 - Political Economy: apply.interfolio.com/188151 All jobs have deadlines Oct 5th. Please help spread the word to potential candidates.

Yiqing Xu@yiqingxu.bsky.social · 4mo ago

1/ Happy to share that "Factorial Difference-in-Differences" (FDID), with Anqi Zhao and @pengding00 , is out in JASA - ACS. doi.org/10.1080/0162... It has been a truly thrilling experience working with Anqi and Peng.

Factorial Difference-in-Differences

We formulate factorial difference-in-differences (FDID), a research design that extends canonical difference-in-differences (DID) to settings in which an event affects all units. In many panel data...

doi.org

Michal Ovádek@movadek.bsky.social · 5mo ago

once again shocked to discover that a complex phenomenon dozens of scientists around the world spend a lot of time and effort studying does not, in fact, neatly reduce to a bivariate plot in the Financial Times. But maybe the next one will

Anton Strezhnev@astrezh.bsky.social · 5mo ago

I don't want to wade too much into this, but the chart going around from that FT birth rates article raised my "making causal claims from time series data" alarm. Compare the figure in the article with the raw TFR data from 1960 onward.

Michal Ovádek@movadek.bsky.social · 5mo ago

My hunch is there is still some low-hanging fruit, the kind of stuff where you burn half a million tokens on a bug that would take 3 seconds of brain time to fix. But the long term economic viability of large-scale AI deployment probably depends on wider, hitherto unrealized efficiency gains.

Ed Zitron@edzitron.com · 5mo ago

AI companies currently subsidize their users $3 to $25 for every dollar of subscription. As a result, users are very sensitive to *any* price or rate limit changes, and have no idea what their actual token burn is. Token-based billing will be lethal at scale. www.wheresyoured.at/premium-what...

So, the current “value” of AI software is mostly driven by the deception of subsidized monthly subscriptions, and even those that have made changes to their business models to improve their margins are still losing money. For example, in October 2025, Augment Code moved everybody to a new “credits” based model from a requests-based one, citing one user that was approaching $15,000 in costs a month on its $250-a-month plan. Even then, Augment Code still sold a money-losing product. Users get 40,000 “credits” that still obfuscate their actual value. Augment gives an example of one task on Claude Opus 4.6 taking up 488 credits, but a “task” could mean anything, and the credit-based system only exists to mask how much you’re actually spending. It turns out the reason is because, per Kilocode, Augment appears to be charging anywhere from 2x to 5x the cost of the actual tokens being burned. The reason that somebody like Augment Code doesn’t want to charge the actual token rates — despite effectively choosing to do so! — is that the moment you go token-based, your users suddenly have to judge your product based on its actual merits. By obfuscating costs using “credits,” Augment Code is attempting to con users into believing they’re getting more for their money than they are. The initial reaction was extremely negative, and now Augment’s Subreddit is a bit of a ghost town, with one user complaining about how arbitrarily Augment burns through credits and another complaining about how buggy the product is.