Many users dismissed advice to review every line of AI-generated code as unnecessary or a sign of low productivity, while others agreed the practice remains essential despite the temptation to skip it.
No Digg Deeper questions have been answered for this story yet.
Most Activity
@yacineMTB no you don't. anyone who has worked on a team knows this
the default state of enterprise codebases are sprawls that no one person on the team has read even 10% of
its easier to do quick loop ups on demand and learn your codebase than maintain a mental model of the entire thing
It's really confusing to me that people think I'm trolling when I say you're supposed to read every line these models produce. It's really important you do. I just shrunk a garbage trash diff of 100 lines to 4 lines
@yacineMTB this is false. once you get to 3k lines of C++ code, you just gotta know how to walk others through your structure but you wouldn’t be able to trace all pointers and functions from memory, you’d have to read them. from memort was always a ridiculous expectation.
@NickADobos @yacineMTB I honestly understand why I am a better engineer than a lot you here. What does it mean “you can’t read all the code”
@yacineMTB I keep creating self contained verifiable tested modules and stack them horizontally and vertically. Never create more than you can verify. Never create monoliths.
Kinda surprising that people don't have the discipline or patience for this. Sucks to suck
@yacineMTB that’s what i just said. dependency free systems have thousands of lines of extra code to compensate.
@epsilver_ don't make it get to 3k lines of cpp code
@decompiled_dev @yacineMTB The ones making money are.
@billmdevs @yacineMTB Honestly maybe I am an idiot and terrible coder lol
But How many words per minute do you read? How many millions of lines is a codebase?
Legitimately if you work on a decently medium to large team there is simply no way to read everything based on time alone.
@yacineMTB i know i know i got carried away
@ojfsaa @yacineMTB i’m stealing this bit
@epsilver_ don't use dependencies
@yacineMTB how do you write any dependency free system without getting to at least 1k?
@NickADobos @yacineMTB Not all codebases are enterprise sprawl
@epsilver_ @yacineMTB just don’t use them, you free up that extra code
@yacineMTB We've replaced coding, not software engineering. Everyone thinks the entire process can now be trusted to a model and they are so wrong.
@yacineMTB ngmi
@yacineMTB Hot take that everyone will hate on: I think its possible to understand the code without reading it. By testing thoroughly end to end. You can still build an accurate mental model.
@yacineMTB I keep reading about people spinning up 3-5 agents and running them in multi-hour loops. How can you possibly do this and still understand every line of code in your codebase?
@yacineMTB You can also ask Claude or Codex to recreate the project from scratch but based on the current implementation while providing clearer guidelines , leaving comments and keeping things organized.