Марина Совина (ночной редактор)
The weary overwinter baker whom I was relieving departed on day three, and from then on, for that first austral summer—November through early February—I was alone every night, the butter thumping against the wall of the bowl in the massive Hobart mixer while I stared out at the flags marking each signatory to the Antarctic Treaty as they bucked in the wind. Headphones in, chef’s jacket on a hook as I peeled down to a tank top, beanie covering my gray-streaked hair, I poked at focaccia, balled cookie dough, frosted cakes, carved up brownies, and cut lemon squares against the background rabble of the tipsy, Catan-obsessed scientists who liked to hang out in the dining room abutting the kitchen.
，这一点在Snipaste - 截图 + 贴图中也有详细论述
The tools used to measure LLM output reinforce the illusion. scc‘s COCOMO model estimates the rewrite at $21.4 million in development cost. The same model values print("hello world") at $19.
SHA512 (FreeBSD-14.4-RELEASE-amd64-ufs.raw.xz) = 9fe2aa5b1988384f5f6306e59b70268defa8ae188e8068b8697b67db2484f19c30639072b54f79a9767f78a02fcee3ab91f8a4fbfae561f14908305174727c8b
About a year ago I wrote AI is a gamechanger for TLA+ users, which argued that AI are a "specification force multiplier". That was written from the perspective an TLA+ expert using these tools. A full 4% of Github TLA+ specs now have the word "Claude" somewhere in them. This is interesting to me, because it suggests there was always an interest in formal methods, people just lacked the skills to do it.