3 ms·

>AI is also great at the mechanics of writing. Should people stop writing? Obviously not. The job of writing is to be read. The job of code is not to be read:

>AI is also great at the mechanics of writing. Should people stop writing? Obviously not. The job of writing is to be read. The job of code is not to be read: it merely helps the humans who write the code. If the code is no longer read by humans, then it no longer needs to be well composed. We want human writers because humans have a better grasp on the ideas and stories that can best move us, and are just better at crafting writing that engages with us. Machines will likely struggle with that for some time, though they can certainly augment and help human writers improve their prose. I get the sadness. I was the one decades ago comparing beautiful code to poetry, and I think that's still true. But also, this type of beauty is craftsmanship that only the craftsman can actually see and appreciate. The reality is that the code just needs to work, and AI can (largely) make it work. Also: the above is assuming that the models continue to improve to human performance, which is of course an open question. I'm not taking a stance here as to whether that will come to pass, merely arguing that it is fundamentally different from writing.

archagon 4mo ago
I fundamentally disagree with this take. With regard to higher-level languages, code is a way to create and compose abstractions around data flow, plus a means of communicating this to other developers (including yourself in the future). Every project brings something new to the table. Machines "solving" this is tantamount to machines developing full-on human intelligence and judgement. The best they can provide is a static facsimile without much hope for future innovation or progress: a snapshot of programming best-ish practices circa 2024 or so. And remember: "just needs to work" includes technical debt, which LLMs are especially great at accruing. When your codebase is composed of 10,000-line, randomly-generated PRs that nobody reads or understands, it's going to be full of untraceable heisenbugs and egregious security holes. What's going to fix them? More 10,000-line PRs? Is it turtles all the way down? Don't mistake my rage at the egregious state of the industry with sadness. In a year, we've thrown much of everything we've learned about solid engineering out the window.

archagon 4mo ago

I fundamentally disagree with this take. With regard to higher-level languages, code is a way to create and compose abstractions around data flow, plus a means of communicating this to other developers (including yourself in the future). Every project brings something new to the table. Machines "solving" this is tantamount to machines developing full-on human intelligence and judgement. The best they can provide is a static facsimile without much hope for future innovation or progress: a snapshot of programming best-ish practices circa 2024 or so. And remember: "just needs to work" includes technical debt, which LLMs are especially great at accruing. When your codebase is composed of 10,000-line, randomly-generated PRs that nobody reads or understands, it's going to be full of untraceable heisenbugs and egregious security holes. What's going to fix them? More 10,000-line PRs? Is it turtles all the way down? Don't mistake my rage at the egregious state of the industry with sadness. In a year, we've thrown much of everything we've learned about solid engineering out the window.