Keyboard Hypnosis
they are the number one users of ai, yet they can’t stop complaining about it.
以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。
they are the number one users of ai, yet they can’t stop complaining about it.
reading the various blog posts, 3 main reasons emerge:
…one might say they look like drug addicts.
“if it’s hurting you so bad, just stop!”.
but of course, as with addictions (i would know), it’s not so simple.
this post is my humble try to find a way out with two words: “just type”.
when you can’t use the brain, use the hands, and the joy of programming will come back.
erik satie has this wonderfully provocative quote:
les pianos, c’est comme les chèques, ça ne fait plaisir qu’à ceux qui les touchent.
an english version would be “(piano) keys are like jackpots, they only please those who hit them”
it’s absolutely true. and code is like a lottery, it only pays off for those who enter it.
people say out loud on conferences: “i didn’t type a single line of code by hand in months”. me neither! and it feels like shit!
for weeks, i tried to reconnect with the joy of programming, and it was the simple joy of typing. of hearing the sounds of the keyboard, of feeling the rebounds of the keys, of trying to match the speed of my thoughts, of lagging behind yet pushing through, somehow.
there are people who still regret keyboard phones, you know!
code is meant to be coded. we all have our favorite language, our favorite “memorized set of instructions that flows from our fingers”. that’s why programmer comes from “pro gramma”, “before the letter”, because we use words that carry more than what they are. they’re instructions, an emerging schema in our brains.
when you type “fix this” in the chat bar, when you’re waiting for claude to answer, when you’re reading in between the typing… you’re losing that flow. that rhythm. that pleasure.
that feeling is yours. it’s yours to retake. whatever claude may say.
replace ai psychosis with keyboard hypnosis.
it doesn’t have to be that antagonistic, though.
there’s none of that in claude code, sure, at least in how we use it now. but that may change.
do you know the size of the context window of the latest models (opus 5.5 and gpt luna, as i type)? in words?
it’s a frikkin 1500-page book. it’s the friggin misérables. it’s three volumes of lord of the rings that you can type as a prompt. why would we shy away?
llms are a way to average your thoughts: if you ramble on for 1500 pages about how you want to build that thing, it will find its way through the mess! put all of your ideas in there, pieces of code, planned developments… no “make a todo app, no mistakes”. no “one-shot”. just the brain dump of a thousand ideas.
maybe that’s the new paradigm. at least it’s a way out. llms are not a tool, not for rich hickey, not for ivan illich, because they use you instead of you using them. but we can find a reverse uno. i know human intelligence can.
maybe in the future the product of a company will be one big ass prompt that they run on each 31 of the month, and they keep that a secret.
in a lot of ways, i’m reminded of this passage in breaking bad: jesse’s wooden box as a testament of his crafty and carefree past.
or that story of a heartbreak one day, and the dad didn’t say anything. the dad offered to spend the weekend building, in the garage, a wooden closet, from saw, planks and nails. and it helped the heartbreak pass.
and that’s for argument 3. code doesn’t have much value now, except a sentimental one.
正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。