As someone that actually reviews software before installing it. No. Error rates are actually up, and lately, error rates on bugfix releases are up by 300% versus 6 months ago.
But, not because I think that vibe coding is bad. I think vibe shipping is bad. It's the laziest shit I have ever encountered in my professional life in software engineering. It basically takes out the entire engineering part, which was already the poorest part in the first place. Go and vibe code. It's awesome. Just... if you wanna ship, spend a couple months on understanding the code you're about to push out there. You have the opportunity now to make REALLY good products. So don't be lazy.
I'll give a real world example: yesterday a bot came up with a critical bug in some code i wrote and suggested a fix. It sounded reasonable and I implemented said fix. Mind you these are highly tuned, optimized and deeply instructed bots with 200-1000 line instructions that are well tested, not garden variety just run claude code / codex and yolo a 3-liner.
You know what? I thought about it and added an extra test fixture. And it broke. Then I dove in. And I found that what sounded extremely plausible and reasonable was in fact a misunderstanding on the part of the bot. If I would have just listened to it and not paid attention, the bug it introduced in the bugfix would have probably gone unnoticed for weeks.
Got it and that definitely makes sense that one double checks whatever they are about to ship more so if it's going to be aimed at a large group of people. Vibe coding shouldn't be an excuse to dump best practices, and I think that's where things get murky sometimes. As for your real life examples that was a good catch and glad you caught it early on. Probably that bug would have been exploited years later like what happend with ColdCard and that would have been really bad for everyone involved. That said, at what point does one draw the line between genuine innovation using vibecoding and the destructive side to it?
Fixed it!
You'll know nothing, but you'll think you're awesome, because Claude said you are
I take it that you are definitely no fan of vibecoding
As someone that actually reviews software before installing it. No. Error rates are actually up, and lately, error rates on bugfix releases are up by 300% versus 6 months ago.
But, not because I think that vibe coding is bad. I think vibe shipping is bad. It's the laziest shit I have ever encountered in my professional life in software engineering. It basically takes out the entire engineering part, which was already the poorest part in the first place. Go and vibe code. It's awesome. Just... if you wanna ship, spend a couple months on understanding the code you're about to push out there. You have the opportunity now to make REALLY good products. So don't be lazy.
I'll give a real world example: yesterday a bot came up with a critical bug in some code i wrote and suggested a fix. It sounded reasonable and I implemented said fix. Mind you these are highly tuned, optimized and deeply instructed bots with 200-1000 line instructions that are well tested, not garden variety just run claude code / codex and yolo a 3-liner.
You know what? I thought about it and added an extra test fixture. And it broke. Then I dove in. And I found that what sounded extremely plausible and reasonable was in fact a misunderstanding on the part of the bot. If I would have just listened to it and not paid attention, the bug it introduced in the bugfix would have probably gone unnoticed for weeks.
Got it and that definitely makes sense that one double checks whatever they are about to ship more so if it's going to be aimed at a large group of people. Vibe coding shouldn't be an excuse to dump best practices, and I think that's where things get murky sometimes. As for your real life examples that was a good catch and glad you caught it early on. Probably that bug would have been exploited years later like what happend with ColdCard and that would have been really bad for everyone involved. That said, at what point does one draw the line between genuine innovation using vibecoding and the destructive side to it?
There is no innovation in vibe coding. It's consolidation, nothing new gets invented if it is not in the prompt. Just a tool. Use it to PoC.