News for hackers. Stuff that matters.
Re:Vibe code is legacy code by mvkel on Wednesday July 30, 2025 @11:05PM (#44741966)
"Tech debt" is a misnomer. It implies that at some point it should be paid down. There are very few examples where dedicating real efforts exclusively to paying down tech debt led to anything beneficial for the customer.

Some of the software we know and love today started with someone writing their very first line of PHP code, or Rails.

Vibe coding is ultimately about getting ideas shipped faster. Code quality is an engineer recruiting/retention strategy.

As we saw with that Tea app (which wasn't even vibe coded), you're only as secure as your weakest firebase store.

Re:Vibe code is legacy code by drysart on Thursday July 31, 2025 @12:28AM (#44742305)
The one tip for quick success that the Human Developers don't want you to know: You don't have to pay down your tech debt if the company folds because hackers leak your customer database and refund all your income.
Re:Vibe code is legacy code by mvkel on Friday August 01, 2025 @10:05PM (#44764351)
This counterfactual implies that successful companies that have not been hacked have done so because they pay down their tech debt.

I can name many companies that have Olympus Mons levels of tech debt, and are doing great. Here's one: Spotify

Re:Vibe code is legacy code by otabdeveloper4 on Thursday July 31, 2025 @03:02AM (#44743139)
> Vibe coding is ultimately about getting ideas shipped faster.

Studies show that using LLMs for coding results in negative net productivity gains on average.

Re:Vibe code is legacy code by lonelyasacloud on Thursday July 31, 2025 @05:48AM (#44744005)
There have been a few studies that show that in certain circumstances this is the case. But doing a literature review will illustrate the bulk support the alternative hypothesis.

Ultimately, whatever the hype factor there is a large number of clever people putting their time and money into it for a reason.

Re:Vibe code is legacy code by mvkel on Friday August 01, 2025 @10:02PM (#44764344)
I would read that study thoroughly.

It had a very specific caveat of "within existing codebases that the engineer was intimately familiar with."

In all other instances, it was not slower