really like to see a source for the first guy's numbers

The citations are inline hyperlinks throughout the piece - IEA, Lawrence Berkeley National Lab, MIT Technology Review, Google's own efficiency data.

They're there. Click them.

Also, neither of those address the effects on creatives' livelihood

We've now gone from "AI can't code," to "AI code is a malicious risk," to "humans would never do that," to "that's a disengenous example" to "ok, but what about this, this and this"

We're verging on a gish gallop at this point, so I demur. Let's stick to the claims at hand instead of litigating shifting goal posts.

On the topic of the second article -

What the judge ruled was that it qualifies as fair use specifically because they destroyed the copies - the destruction is what made the scanning fair use, not what made it wrongful.

Retaining the digital files without destroying the physical copies would have been the violation.

Meaning the law perversely incentivised this behavior. (If you want the actual court reporting, the Ars Technica piece the article quotes is the cleaner source).

https://arstechnica.com/ai/2025/06/anthropic-destroyed-millions-of-print-books-to-build-its-ai-models/

So, it's more complicated than "Anthropic are book burners."