This is highly debatable. They look better, but whenever I inspect the walls and see what is holding the house together, I just shrug and leave my contact info, because I know I’ll be called to fix the mess.
> ...and building them increasingly feels like magic.
Engineering should not feel like magic to those with a proper education. Otherwise, we are merely practicing alchemy.
Maybe that was possible 2 years ago, but for the last year it's just not simply true.
Sure there will be codebases that will get messy, outgrow the amateur vibe coder, etc. But AI is already better than 90% coders. Certainly can focus better than 99% of coders.
> Every file had a comment at the top with the date and the author’s name.
Yeah that is most likely less about pride in the craftsmanship and more likely a reflex from bad source version control software.
EDIT: Also, you know they didnt type these comments by hand right? And that the first code completion systems date back to the 70s, granted it was mostly spell checkers.
I joined Asobo Studio[1] for an internship back in 2010, and some code there was already nearly 15 years old as they grabbed the remains of Kalisto[2] including the engine/editor. That editor (Pawap) is still used to this day, although the engine (Zouna) has been rewritten (but uses the same data backend as the old code, albeit heavily updated too). This was probably the oldest codebase I worked on at the time.
I have an active project where parts of it still uses on the back-end WebObjects, an ORM framework originally from NeXT and later apple. Over the years parts of it are being rewritten (WebObjects is not threadsafe) but it just keeps running and running processing millions of transactions per week.
As part of my current assignment I see code and tests that are 15 years old (or more), and I only hope that people who wrote this code have become better as software engineers since then. Otherwise there is no hope for humanity.
macOS's peak was roughly 15 years ago, to be fair. Apple hasn't ever gotten things fully right after Snow Leopard. The 15 year old code is probably the best code in the stack.
The irony of this metaphor is that new housing construction these days is really bad, just watch any of the inspectors posting to youtube or tiktok. Even those new 3d printed houses are crumbling after a year, and they can't be fixed because of the way they're built.
We had better built houses when they came as kits from Sears than we have today.
I agree with Mitchell's original post: "stay frosty with AI", and I disagree with response that says "the houses are better". I'm assuming that Mitchell was saying you should treat all AI code with caution, and the more I work with AI-generated code, the more I agree with this sentiment. Even the frontier models (Fable, Opus, Sol, etc) all make fundamental errors from time to time, and if you're not keeping your eyes open these are easy to miss.
Just the other day, I got Fable to help generate a sign-up flow for a Django app. It looked perfect to me, and like the Twitter post linked to, I looked at it with marvel and delight. But I got Sol to review the code and it spotted some fundamental issues with the flow that would have let bad actors enumerate accounts and log in before their account was verified - both of which were supposed to be checked.
To continue with the analogy, yes we can "will houses into existence", but unless we've checked the bricks and the foundations, these houses can easily crumble into rubble.
22 comments
[ 0.78 ms ] story [ 12.7 ms ] threadThis is highly debatable. They look better, but whenever I inspect the walls and see what is holding the house together, I just shrug and leave my contact info, because I know I’ll be called to fix the mess.
> ...and building them increasingly feels like magic.
Engineering should not feel like magic to those with a proper education. Otherwise, we are merely practicing alchemy.
Maybe that was possible 2 years ago, but for the last year it's just not simply true.
Sure there will be codebases that will get messy, outgrow the amateur vibe coder, etc. But AI is already better than 90% coders. Certainly can focus better than 99% of coders.
Yeah that is most likely less about pride in the craftsmanship and more likely a reflex from bad source version control software.
EDIT: Also, you know they didnt type these comments by hand right? And that the first code completion systems date back to the 70s, granted it was mostly spell checkers.
15 years ago people weren't crafting artisanal lines of code to build by dry-stacking.
I bet some of the oldest artifacts in code say "TODO" or "Temporary work-around"...
[1] https://en.wikipedia.org/wiki/Asobo_Studio [2] https://en.wikipedia.org/wiki/Kalisto_Entertainment
Are they? Why? Even in the brick metaphor that doesn't work.
It's windows - its broken.
I worked on a reactor physics simulation app with 50 year old code then, which is now approaching 75 years old now. It compiled on Windows and *NIX.
We had better built houses when they came as kits from Sears than we have today.
Just the other day, I got Fable to help generate a sign-up flow for a Django app. It looked perfect to me, and like the Twitter post linked to, I looked at it with marvel and delight. But I got Sol to review the code and it spotted some fundamental issues with the flow that would have let bad actors enumerate accounts and log in before their account was verified - both of which were supposed to be checked.
To continue with the analogy, yes we can "will houses into existence", but unless we've checked the bricks and the foundations, these houses can easily crumble into rubble.