[znpy]

> I understand why it wouldn’t be feasible for a human to do this, but I’m quite sceptical about an AI assessing how accurate predictions turned out to be/how contrarian they were at the time.

Quite the contrary actually. The economist has a fairly consistent editing style that gets enforced, so its writing style is very linear and very consistent, fairly easy to “understand” for and advanced llm like GPT-5.5

[flashman]

You assert that, but can you prove it? Likewise I can say that the model is biased towards rating the Economist's claims as 'mainstream' and 'bang on' because as a high-profile publication those claims are repeated elsewhere and hence up-weighted.

I could be wrong - my point is, how would we know?

[znpy]

https://news.ycombinator.com/item?id=26700664

> The reason why the Economist articles all read the same is that they go through a process called "subbing" or sub-editing by the same small group of editors who own the Economist's "voice".

> I interned at the Economist one summer in college and wrote two articles for it - the published articles bore a small relationship to what I had submitted. The sub-editors seemed to be mostly George Smiley-type Oxford and Cambridge PhD's of a certain age with an eclectic range of expertises, as far as I recall.