Is all vibe coding actually like that too or does some of it actually work?

It’s all like that.

How bad that is - for you - depends on your patience and your learning style.

When I use it, my experience usually lets me recognize the mistakes and correct them quickly. So it’s just a lazy convenience. Most of the time.

I’ve had it make subtle mistakes that cost me significant amounts of time to cleanup after letting the vibe code run for a few minutes.

I’m aware that particular mistake cost me more time that vibe coding has ever saved me.

I don’t mind, because my employer is excited about AI right now, and I get paid for my time, and I don’t work unpaid overtime.

So - to your implied questions:

Is AI bad at coding?

Yes. It will get better. But today, it is worse than most people think. Obvious problems are easily fixed. Subtle problems are being released daily all over the Internet to combine to cause headaches later.

Should you try it, anyway?

Of course! You’ll learn something and it might do a good enough job for what you need. If you stick with it, you’ll learn enough to do what you need.

Is vibe coding a better path forward than learning a programming language?

Absolutely not. If you need to succeed, and had to pick one, learn to code.

But you don’t have to pick just one approach. And it’s probably impossible to vibe code for long without learning to actually code. Vibe coding is a path toward aware knowledgeable coding. It’s not the only path. It’s not the best path. But it’s still a path. And you can pursue more than one path.

So I say, Dive in! You’ll be complaining with the rest of us, soon! Maybe together we will make it a bit better.

Full complex app, forget about it. It’s not going to work.

Concrete functions or small parts of the program (or maybe a very small project? 50/50 chance. Depending on the complexity.

For instance I benchmark several LLMs last month, asking them to build a worldle clone for the terminal. Some of them were able to spite the full program in a completely working state.

For anything larger or more complex I haven’t had any luck. And LLM are mostly used for references and ideas.

You know those movies where the guy gets 3 wishes from a Genie who takes malicious delight in giving them exactly what they asked for even when they’re super careful like “I want a million dollars, and no I don’t want it stolen from a bank, or anywhere that someone’s going to come after me for having it and oh, it needs to be actual real US dollars in circulation today, and without any tax obligations, the IRS can’t come after me. The SEC can’t come after me.” And when they think finally that they’ve specified everything they possibly can, the Genie summons the money and a big gust of wind blows it all out the window and down the street… Then they need to use their second wish to summon it all back in and shut the window. But then the genie summons it back into the fireplace and it all catches fire, so they have to use their third wish to bring it all out of the fireplace, so the Genie brings it all out, but it’s just ashes…

Well, okay, there’s probably no movie like that, but that’s what programming with AI is like.

“Vibe coding” purists define it as “If you know how it works then it isn’t vibe coded”. And those type of coders kinda keep going at it more and more refined until they eventually get some spaghetti code that kinda does what they wanted it to do and heck, It’s close enough, ship it! Then they end up being exploited by some random internet hacker.

Most of the companies that use “Agentic coding” are using it to perform rapid prototyping or templating, performing repetitive tasks quickly or generally using it like a really dumb junior programmer, that the engineer then takes their code and does the code review / testing (often again using AI tools), followed by a whole heap of fixing up, to make sure it does what it says on the box.

As stated on other comments, the amounts of money they pay for this kind of AI tooling could easily cost many thousands of dollars a month (in addition to the engineer(s) salary/salaries), but the order of magnitudes of productivity increase for that engineer make it worthwhile. But you need that experienced engineer to make it all work.

I’m not aware of any companies that are solely using coding agents in isolation to replace engineers completely. I’m sure it’ll happen one day and I’ll probably be forced into retirement at that point.

I don’t even have to read everything you wrote past the question.

no. no it does not.

it doesn’t work for many reasons. most of all it doesn’t work when you need to improve or extend the code. handing it over to a new developer also doesn’t work.

If I ever see another developer vibe code IRL I will relentlessly mock them until HR is forced to get involved.

Here’s Simon Willison’s write up of how he uses AI. He’s been using it for a couple of years and distilled his methods in this article. He also discussed when and how he vibe codes.

LLMs are great at language problems. If you’re learning the syntax of a new programming language or you’ve forgotten the syntax for a specific feature, LLMs will give you exactly what you want.

I frequently use AI/LLMs when switching languages to quickly get me back up to speed. They’re also adequate at giving you a starting point, or a basic understanding of a library or feature.

The major downfall is if you ask for a solution to a problem. Chances are, it will give you a solution. Often it won’t work at all.
The real problem is when it does work.

I was looking for a datatype that could act as a cache (forget the oldest item when adding a new one). I got a beautifully written class with 2 fields and 3 methods.
After poking at the AI for a while, it realized that half the code wasn’t actually needed. After much more prodding, it finally informed me that there was actually an existing datatype (LinkedHashMap) that would do exactly what I wanted.

Be aware that AI/LLMs will rarely give you the best solution, and often give you really bad solutions even when an elegant one exists. Use them to learn if you want, but don’t trust them.

My bro, your TA wasn’t better at coding because “higher IQ”. They were better because they put in the hours to build the instincts and techniques that characterize an experienced developer. As for LLM usage, my advice is to be aware of what they are and what the aren’t. They are a randomized word prediction engine trained on— among other things— all the publicly available code on the internet. This means they’ll be pretty good at solving problems that it has seen in its training set. You could use it to get things set up and maybe get something partway done, depending on how novel your idea is. An LLM cannot think or solve novel problems, and they also generally will confidently fake an answer rather than say they don’t know something, because truly, they don’t know anything. To actually make it to the finish line, you’ll almost certainly need to know how to finish it yourself, or learn how to as you go.

I vibe coded an AutoHotKey script to automate part of my job. It works.

The exact definition of vibe coding varies with who you talk to. A software dev friend of mine uses ChatGPt every day in his work and claims it saves him a ton of time. He mostly does db work and node apps right now, and I’m pretty sure the way he uses ChatGPT falls under the heading of vibe coding - using AI to generate code and then going through the code and tweaking it, saving the developer a lot of typing and grunt work.

if you know how to code, you can vibe code because you can immediately see and be confident enough to identify and not use obvious mistakes, oversights, lack of security, and missed edge cases the LLM generated.

if you don’t know how to code, you can’t vibe code, because you think the LLM is smarter than you and you trust it.

Imagine saying “I’m a mathematician” because you have a scientific calculator. If you don’t know the difference between RAD and DEG and you just start doing calculations without understanding the unit circle, then building a bridge based on your math, you’re gonna have a bad time.

Think of LLMs as the person who gets good marks in exams because they memorized the entire textbook.

For small, quick problems you can rely on them (“Hey, what’s the syntax for using rsync between two remote servers?”) but the moment the problem is slightly complicated, they will fail because they don’t actually understand what they have learnt. If the answer is not present in the original textbook, they fail.

Now, if you are aware of the source material or if you are decently proficient in coding, you can check their incorrect response, correct it, and make it your own. Instead of creating the solution from scratch, LLMs can give you a push in the right direction. However, DON’T consider their output as the gospel truth. LLMs can augment good coders, but it can lead poor coders astray.

This is not something specific to LLMs; if you don’t know how to use Stackoverflow, you can use the wrong solution from the list of given solutions. You need to be technically proficient to even understand which one of the solutions is correct for your usecase. Having a strong base will help you in the long run.

Vibe coding works, but there are some serious caveats.

I’ve used LLMs for data visualization and found them helpful for simple tasks, but they will always make serious mistakes with more complex prompts. While they understand syntax and functions well, they usually produce errors that require manual debugging. Vibe coding with LLMs works best if you’re an expert in your project and could write all of the code yourself but just can’t be bothered. Prepare to spend some time fixing the bugs, but it should still be faster than writing all of it yourself.

If you’re not proficient in using a specific function the LLM generated, vibe coding becomes less effective because debugging can be time consuming. Relying on an LLM to troubleshoot its own code tends to lead to “fixes” that only spawn more errors. The key is to catch these situations early and avoid getting lured into any of the wild goose chases it offers.

In my experience, an LLM can write small, basic scripts or equally small and isolated bits of logic. It can also do some basic boilerplate work and write nearly functional unit tests. Anything else and it’s hopeless.

I have successfully written and deployed a number of large complex applications with 100% AI written code, but I micromanage it. I’ve been developing software for 30 years and use AI as a sort of code paintbrush. The trick is managing the AI context window to keep it big enough to understand its task but small enough to not confuse it.

I use it as a friendlier version of stackoverflow. I think you should generally know / understand what you are doing because you have to take everything it says with a grain of salt. It’s important to understand that these assistants can’t admit that they don’t know something and come up with random generated bullshit instead so you can’t fully trust their answers.

So you still need to understand the basics of software development and potential issues otherwise it’s just negligence.

On a general note: IQ means nothing. I mean a lot of IQ tests use pattern recognition tasks that can be helpful but still, having a high IQ says nothing about you ability as developer