Yup. In this era of rage bait and misinformation it's better to ignore it and post something nice instead.

I don't read a lot of history either, but really enjoyed Unruly by David Mitchell. I'll add this book to the list.

Sure, you can say it's all so advanced it's magic, but that comes at the expense of any tension or relatability. It also happens to be much easier to write for.

I think the even more stupider decision was the doubling down on fantasy over SciFi. The fact that the writers could show Piper at the end is because they've given up on any form of logic long ago.

The data probably won't be that dynamic though, so it's really a matter of multiple projections for different query dimensions with appropriate cluster configuration managed with something like DBT. Maybe good enough?

It was when we collectively realized that *we* were not Google's target market.

That was the point when lines on the graph crossed and DC overtook Marvel.

It took me 30 odd years, but these little color-coded rubber cable ties are a game changer for the cable draw. a.co/d/j9nsuZs

And this, which debunks the whole thing. open.substack.com/pub/cybersec...

Just reading tedium.co/2025/09/23/s...

So X is an example of model collapse from synthetic (bot) data? Yeah, holds up.

I don't think it has done well. I think it's out of the top 10 already over here. I wonder how much better a more literal version would have gone down.

The problem for me was (assumingly) the studio decision to hire Columbus and instruct him to do a Harry Potter job on the book.

Maybe a good data architecture can avoid the need for partial functions. For example, verifying customer id by asking for other verifiable data in the tool function like name. Joining and projection denormalized datasets from multiple sources and exposing a single MCP also helps a lot.