smitop

The post's explanation for the content filtering policy 400 error isn't correct. Anthropic doesn't use that error for safety reasons ("These refusals do not reflect Anthropic’s judgments about the propriety of any content"[1]); content blocked by classifiers get a different kind of error.

The content filtering 400 is only for copyright blocks, where Anthropic detects when Claude is outputting copyrighted text and blocks it to prevent copyright infringement. The book in question is in the public domain in the US (but not all countries) as of January 1, 2026; Anthropic probably just doesn't automatically remove works from the copyright filter when their copyright expires.

[1]: https://privacy.claude.com/en/articles/10023638-why-am-i-rec...

I don't think this is really a problem; I wrote about my thoughts at https://iter.ca/post/git-malleate/

Actions runs from external contributors aren't run with Actions secrets; if you are using Actions right (i.e. not using pull_request_target wrong) you don't need to trust external contributors. (eta: iirc the original point of the Actions approval flow was preventing cryptomining spam from abusing free compute)

There are other stablecoins that aren't scams though, like USDC. I think Stripe would probably either create their own USD stable or partner with Circle.

> if you fail to mention that something was made by AI as part of a compound work you could end up losing copyright over the whole thing

The source you linked says the opposite of that: "the inclusion of elements of AI-generated content in a larger human-authored work does not affect the copyrightability of the larger human-authored work as a whole"

YouTube is currently running an A/B test for server-side insertion according to what some other people have posted. I'm not getting SSAI ads so I can't really know much about them though.

Most people searching for "elm" are researching elm trees, not the language - the highest search activity was in May 2010, before Elm existed.

Insurance companies could front the money, in exchange for a premium based on how legitimate the claim seems.

tflite files can contain a ResizeOp that resizes the image: https://ai.google.dev/edge/api/tflite/java/org/tensorflow/li...

The file is only 7.7kb, so it couldn't contain many weights anyways.

Your math is wrong - it's $4B ÷ 815k residents ÷ $27k = 0.18 used Tesla Model 3s per resident. (you accidentally used 27k as both the price and the population)

JSTOR doesn't get any tax benefit for doing this because they're already fully exempt from US income tax - their parent, Ithaka, is a 501(c)(3) non-profit (see https://en.wikipedia.org/wiki/Ithaka_Harbors).

Author here, I removed the background.

Grey market phones aren't stolen. Samsung sells phones cheaper in some regions, so arbitrageurs buy them cheap in one region and resell them at a higher price in Mexico. Samsung wouldn't sell region-unlocked phones at the cheaper price because they want to make money.

Ah fuck, I grepped the YouTube transcript for WASM/WebAssembly but I guess youtube's automatic captions aren't very good. I feel silly now for not watching the video before commenting.

fun trivia: in WebAssembly it's valid to read/write to memory address 0.