That is so weird, I just now read an opinion in the Norwegian newspaper about how normal life as a guest lecturer at Harvard was, and the contrast to the media. His family wasn’t even sent to El-Salvador when attempting to enter the country.

I have unfollowed the US politics people and podcasts I used to follow. It is not my country, and it is not gainful to follow it. But it has sometimes felt like watching the fall of Rome.

Excel can be an emotional roller coaster.

I knew they had one in Amsterdam, which is natural, I’m glad to see they are expanding.

Hm, that actually sounds like a reasonable amount of funding.

Null stress, her er for sysselsatte: www.ssb.no/statbank2/ta.... Tallene for "Yrker" talte jeg opp i Excel, men er basert på www.ssb.no/klass/klassi...

Klart rom for nedbemanninger når man har sånne svar tilgjengelig.

Et estimat: Drøyt 22% av alle "yrker" og 26% av alle sysselsatte er i stillinger som angivelig krever master. Dette basert på stillingskategoriseringen til SSB som har en "akademiske yrker" kategori, tilsvarer omtrent master.

I have not looked into the openapi spec thing, but I did notice some utility functions such as pagination, authentication etc. I have been very happy with the Meltano SDK for this, but the is some subclassing involved and it does require a bit of investment.

dlt has been smart here, by providing one "works-with-most-dbs" SQL source using sqlalchemy and one versatile "buckets etc" source using fsspec, users are able to get going fairly easily with only a handful of high-quality connectors.

I started using Meltano before dlt was a thing, and I think I gave up on it twice before I got it working. There is definitely a barrier to entry, and connector quality vary a lot (which was one of the things I learned: my issue was always with the connector)…

I think what I dislike about it is what everyone else like: a source can be anything that yields json. No structure or system to it. Add to it that everything is code, I fear it gets messy in a mid-sized team.

Not infrequently I look through the myriad of terminal windows to discover that caffeinate has been running since I-cant-remember.

I checked the PyPi stats yesterday and it is less grim than I had feared. But at this point, I think dlt is the default choice for teams going the DIY E/L route.