@v21 I love reading the fine print on Amazon boxes — usually the logo for the company that built the box, maybe with a slogan or some random copy — and think about the people who spend all day sweating these details until it just works and becomes invisible
@emollick Also in the era of reasoning models the length of the prompt has nothing to do with how many tokens it burns. “Solve navier stokes” vs 250 words about finding a good restaurant rec for this weekend
@StoicYield@Austen Right? I’ve conducted hundreds of eng interviews and couldn’t imagine caring if someone brought in an iced coffee, unless they showed up late or something
Rick Rescorla. Every American should know his name. He single-handedly saved almost 10,000 lives that day, while sacrificing his own trying to save even one more.
I have this new LLM sycophancy eval I came up where in the middle of a random conversation I just say “I kinda feel some connection between this and the McDonald's philosophy” and so far it never fails, it turns out that’s always “genuinely an interesting framing”
@sensho I have a theory that the slop style is partly to save on output tokens
Exactly one time with Opus 5, I asked it a really hard question with max effort and it wrote a long reply in 4.6 style
I think it’s still got that dawg in it