More weed use among teens → more hot dogs sold in the U.S.

Just for fun, this was Opus’s attempt at translation with the prompt: “I saw this in a commercial for tea. Please translate: 「美味ちい」Keep an eye out for wordplay or cultural difference that would matter in the translation.”

This is clever Japanese wordplay! 「美味ちい」 is a pun that combines:

• 美味しい (oishii) = "delicious"
• 茶 (cha) = “tea”

They’ve replaced the し (shi) in “oishii” with ち (chi), creating a portmanteau that sounds like “oichii.” This blends “delicious” with a phonetic hint at “tea” (since ち/chi is close to ちゃ/cha).

Translation: “Deli-tea-ous” or “Tea-licious”

In English, we’d need a similar pun to capture the playful spirit. The commercial is essentially making a tea-specific version of the word “delicious” - it’s a fun bit of marketing language that suggests the tea is so good it deserves its own special adjective!

Interestingly, Anthropic’s newer (but smaller) model, Sonnet 4.5, doesn’t catch on without following up with something to the effect of, “What is tea in Japanese?” LLMs showed pretty amazing multilingual capabilities before they even started intentionally training them with more languages. That said, you definitively still need someone who speaks both languages to check/guide em. As it is their way to bullshit.

Been in video production for 30 years. We cannot rely on spotting problems to identify AI video. Selective attention alone means we miss major details in traditional video. Streaming services are testing AI upscaling, adding AI artifacts to pre-AI media. I regularly see comments pointing out artifacts from other sources (interlacing, morph cuts, moire, stabilization, compression, data moshing, etc) as evidence that the video is AI generated.

I’ve heard people say that using AI image generators is a technical skill. It’s not. It’s about as hard as using Google.

Ever watch other people search? It’s definitely a skill.

The Yale study

I get that too. But many people who are going to do harm won’t care what you do to ingratiate yourself. They’re more likely to see it as a sign of weakness that makes an easier target. We can’t service our way to safety. And on an even more fundamental level, the never-ending grasping for security through impression management is a way to distract us from the fear of death. “Memento mori,” in part is meant to short circuit this drive, which can turn malignantly toward sacrificing humanity and freedom for illusions of immortality and invincibility.

Probably not, but this is what I try to remember. People pleasing is a weird kind of arrogance. It feels like a selfless sacrifice, but it’s fundamentally an attempt at control. We cannot control the internal state of others. People are not going to always be pleased, nor should they be. And on the other side, those that encourage people pleasing behavior are (perhaps unintentionally) trying to control or exploit me.

That’s not what’s meant by “ears over shoulders,” rather they’re over when standing. So in-line with shoulders when laying down, not what’s pictured here. But whether or not you want the neck in flection depends on whether or not there’s been an injury. The kinetic chain isn’t simple.