Fair: it was rude. Moderation is hard and I respect what you do. But it's also a sentiment several other comments expressed. It's the conversation we're having. Can we have any discussions of code quality without making assumptions about each others' code quality? I mean, yeah, I could probably have done better.
Oh for sure you can talk about this, it's just a question of how you do it. I'd say the key thing is to actively guard against coming across as personal. To do that is not so easy, because most of us underestimate the provocation in our own comments and overestimate the provocation in others (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...). This bias is like carbon monoxide - you can't really tell it's affecting you (I don't mean you personally, of course—I mean all of us), so it needs to be consciously compensated for.
As for those other comments - I take your point! I by no means meant to pick on you specifically; I just didn't see those. It's pretty random what we do and don't see.
I understand the provocation, but please don't respond to a bad comment by breaking the site guidelines yourself. That only makes things worse.
Your GP comment was great, and probably the thing to do with a supercilious reply is just not bother responding (easier said than done of course). You can usually trust other users to assess the thread fairly (e.g.https://news.ycombinator.com/newsguidelines.html
> What makes you think I'm not "a developer who strongly values brevity and clarity"
Some pieces of evidence that make me think that:
1. The base rate of developers who write massively overly verbose code is about 99%, and there's not a ton of signal to deviate from that base rate other than the fact that you post on HN (probably a mild positive signal).
2. An LLM writes 80% of your code now, and my prior on LLM code output is that it's on par with a forgetful junior dev who writes very verbose code.
3. 200K lines of code is a lot. It just is. Again, without more signal, it's hard to deviate from the base rate of what 200K-line codebases look like in the wild. 99.5% of them are spaghettified messes with tons of copy-pasting and redundancy and code-by-numbers scaffolded code (and now, LLM output).
This is the state of software today. Keep in mind the bad programmers who make verbose spaghettified messes are completely convinced they're code-ninja geniuses; perhaps even more so than those who write clean and elegant code. You're allowed to write me off as an internet rando who doesn't know you, of course. To me, you're not you, you're every programmer who writes a 200k LOC B2B SaaS application and uses an LLM for 80% of their code, and the vast, vast majority of those people are -- well, not people who share my values. Not people who can code cleanly, concisely, and elegantly. You're a unicorn; cool beans.
Before you used LLMs, how often were you copy/pasting blocks of code (more than 1 line)? How often were you using "scaffolds" to create baseline codefiles that you then modified? How often were you copy/pasting code from Stack Overflow and other sources?
> I'm struggling to even describe... 200,000 lines of code is so much.
The point about increasing levels of abstractions is a really good one, and it's worth considering whether any new code that's added is entirely new functionality, some kind of abstraction over some existing functionality (that might then reduce the need for as new code), or (for good or bad reason) some kind of copy of some of the existing behaviour but re-purposed for a different use case.
200kloc is what, 4 reams of paper, double sided? So, 10% of that famous Margaret Hamilton picture (which is roughly "two spaceships worth of flight code".) I'm not sure the intuition that gives you is good but at least it slots the raw amount in as "big but not crazy big" (the "9 years work" rather than "weekend project" measurement elsethread also helps with that.)
It is quite a putdown to tell someone else that if you wrote their program it would be 10 times shorter.
That's not in keeping with either the spirit of this site or its rules: https://news.ycombinator.com/newsguidelines.html.
Fair: it was rude. Moderation is hard and I respect what you do. But it's also a sentiment several other comments expressed. It's the conversation we're having. Can we have any discussions of code quality without making assumptions about each others' code quality? I mean, yeah, I could probably have done better.
> "That would probably be 1000 line of Common Lisp." https://news.ycombinator.com/item?id=44974495
> "Perhaps the issue is you were used to writing 200k lines of code. Most engineers would be agast at that." https://news.ycombinator.com/item?id=44976328
Oh for sure you can talk about this, it's just a question of how you do it. I'd say the key thing is to actively guard against coming across as personal. To do that is not so easy, because most of us underestimate the provocation in our own comments and overestimate the provocation in others (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...). This bias is like carbon monoxide - you can't really tell it's affecting you (I don't mean you personally, of course—I mean all of us), so it needs to be consciously compensated for.
As for those other comments - I take your point! I by no means meant to pick on you specifically; I just didn't see those. It's pretty random what we do and don't see.
[flagged]
I understand the provocation, but please don't respond to a bad comment by breaking the site guidelines yourself. That only makes things worse.
Your GP comment was great, and probably the thing to do with a supercilious reply is just not bother responding (easier said than done of course). You can usually trust other users to assess the thread fairly (e.g.https://news.ycombinator.com/newsguidelines.html
> What makes you think I'm not "a developer who strongly values brevity and clarity"
Some pieces of evidence that make me think that:
1. The base rate of developers who write massively overly verbose code is about 99%, and there's not a ton of signal to deviate from that base rate other than the fact that you post on HN (probably a mild positive signal).
2. An LLM writes 80% of your code now, and my prior on LLM code output is that it's on par with a forgetful junior dev who writes very verbose code.
3. 200K lines of code is a lot. It just is. Again, without more signal, it's hard to deviate from the base rate of what 200K-line codebases look like in the wild. 99.5% of them are spaghettified messes with tons of copy-pasting and redundancy and code-by-numbers scaffolded code (and now, LLM output).
This is the state of software today. Keep in mind the bad programmers who make verbose spaghettified messes are completely convinced they're code-ninja geniuses; perhaps even more so than those who write clean and elegant code. You're allowed to write me off as an internet rando who doesn't know you, of course. To me, you're not you, you're every programmer who writes a 200k LOC B2B SaaS application and uses an LLM for 80% of their code, and the vast, vast majority of those people are -- well, not people who share my values. Not people who can code cleanly, concisely, and elegantly. You're a unicorn; cool beans.
Before you used LLMs, how often were you copy/pasting blocks of code (more than 1 line)? How often were you using "scaffolds" to create baseline codefiles that you then modified? How often were you copy/pasting code from Stack Overflow and other sources?
At least to me what you said sounded like 200k is just with LLMs but before agents. But it's a very reasonable amount of code for 9 years of work.
This is such a bizarre comment. You have no idea what code base they are talking about, their skill level, or anything.
> I'm struggling to even describe... 200,000 lines of code is so much.
The point about increasing levels of abstractions is a really good one, and it's worth considering whether any new code that's added is entirely new functionality, some kind of abstraction over some existing functionality (that might then reduce the need for as new code), or (for good or bad reason) some kind of copy of some of the existing behaviour but re-purposed for a different use case.
200kloc is what, 4 reams of paper, double sided? So, 10% of that famous Margaret Hamilton picture (which is roughly "two spaceships worth of flight code".) I'm not sure the intuition that gives you is good but at least it slots the raw amount in as "big but not crazy big" (the "9 years work" rather than "weekend project" measurement elsethread also helps with that.)