[Deestan]

I am now making an emotional reaction based on zero knowledge of the B2B codebase's environment, but to be honest I think it is relevant to the discussion on why people are "worlds apart".

200k lines of code is a failure state. At this point you have lost control and can only make changes to the codebase through immense effort, and not at a tolerable pace.

Agentic code writers are good at giving you this size of mess and at helping to shovel stuff around to make changes that are hard for humans due to the unusable state of the codebase.

If overgrown barely manageble codebases are all a person's ever known and they think it's normal that changes are hard and time-consuming and needing reams of code, I understand that they believe AI agents are useful as code writers. I think they do not have the foundation to tell mediocre from good code.

I am extremely aware of the judgemental hubris of this comment. I'd not normally huff my own farts in public this obnoxiously, but I honestly feel it is useful for the "AI hater vs AI sucker" discussion to be honest about this type of emotion.

[mind-blight]

It really depends on what your use case is. E.g. of you're dealing with a lot of legacy integrations, dealing with all the edge cases can require a lot of code that you can't refactor away through cleverness.

Each integration is hopefully only a few thousand lines of code, but if you have 50 integrations you can easily break 100k loc just dealing with those. They just need to be encapsulated well so that the integration cruft is isolated from the core business logic, and they become relatively simple to reason about

[bubblyworld]

> 200k lines of code is a failure state.

What on earth are you talking about? This is unavoidable for many use-cases, especially ones that involve interacting with the real world in complex ways. It's hardly a marker of failure (or success, for that matter) on its own.

[haskellshill]

If all your code depends on all your other code, yeah 200k lines might be a lot. But if you actually know how to code, I fail to understand why 200k lines (or any number) of properly encapsulated well-written code would be a problem.

Further, if you yourself don't understand the code, how can you verify that using LLMs to make major sweeping changes, doesn't mess anything up, given that they are notorious for making random errors?

[throwawaymaths]

200k loc is not a failure state. suppose your b2b saas has 5 user types and 5 downstream SAASes it connects to, thats 20k loc per major programming unit. not so bad.

[krainboltgreene]

That's actually insane.

[johnnyanmac]

I agree on principle, and I'm sure many of us know how much of a pain it is to work on million or even billion dollar codebases, where even small changes can be weeks of beauracracy and hours of meetings.

But with the way the industry is, I'm also not remotely surprised. We have people come and go as they are poached, burned out, or simply life circumstances. The training for the new people isn't the best, and the documentation for any but the large companies are probably a mess. We also don't tend to encourage periods to focus on properly addressing tech debt, but focusing on delivering features. I don't know how such an environment over years, decades doesn't generate so much redundant, clashing, and quirky interactions. The culture doesn't allow much alternative.

And of course, I hope even the most devout AI evangelists realize that AI will only multiply this culture. Code that no one may even truly understand, but "it works". I don't know if even Silicon Valley (2014) could have made a parody more shocking than the reality this will yield.