Suhas Guruprasad

Against vibe coding

They kept trying to mess up the style. They're bloating complexity. It's a total mess. It's not net useful. It's annoying to have to type out what I want in English.

Andrej Karpathy in his interview with Dwarkesh Patel https://www.youtube.com/watch?v=lXUZvyajciY

This is a curated bunch of recent internets blog-sphere writing against vibe coding. Most of them made it to HN front page at some point and I just collected them in my bookmarks because I simply thought it would be good to collect such articles and no other reason. The links are in order of me discovering / stumbling upon them. Btw, this post does not mean that I'm personally against vibe coding. I am just collecting these for academic purposes.

  1. Roger Goldfinger in his post "Claude Code is a Slot Machine" https://rgoldfinger.com/blog/2025-07-26-claude-code-is-a-slot-machine/

Claude Code keeps me waiting. Here I am pressing return like a crack-addicted rodent in a lab. “Yes, I want to make this edit.” I watch as it works, glassy-eyed and bored as the code scrolls by, and on the edge of my seat because my ideas are about to become reality.

  1. Jason Gorman in his post "Comprehension Debt" https://codemanship.wordpress.com/2025/09/30/comprehension-debt-the-ticking-time-bomb-of-llm-generated-code/

What is new is the scale of the problem being created as lightning-speed code generators spew reams of unread code into millions of projects.

  1. Vincent Quigley in his post "First attempt will be 95% garbage" https://www.sanity.io/blog/first-attempt-will-be-95-garbage

First attempt (95% garbage rate). Claude builds context about your system. You identify the actual challenges. The code is usually completely wrong.

  1. Alex Kondov in his post "I Know When You're Vibe Coding" https://alexkondov.com/i-know-when-youre-vibe-coding/

No one would implement a bunch of utility functions that we already have in a different module. No one would change a global configuration when there’s a mechanism to do it on a module level. No one would write a class when we’re using a functional approach everywhere.

  1. Donado Lab's "Vibe Coding Cleanup As A Service" https://donado.co/en/articles/2025-09-16-vibe-coding-cleanup-as-a-service/

Hamid Siddiqi manages 15-20 cleanup projects simultaneously, charging premium rates to untangle what he calls "AI spaghetti" - inconsistent interfaces, redundant functions, and business logic that makes no sense.

  1. Sean Goedecke in "What is "good taste" in software engineering?" https://www.seangoedecke.com/taste/

Technical taste is different from technical skill. You can be technically strong but have bad taste, or technically weak with good taste. Like taste in general, technical taste sometimes runs ahead of your ability: just like you can tell good food from bad without being able to cook, you can know what kind of software you like before you’ve got the ability to build it. You can develop technical ability by study and repetition, but good taste is developed in a more mysterious way.

  1. Kix Panganiban in "Two things LLM coding agents are still bad at" https://kix.dev/two-things-llm-coding-agents-are-still-bad-at/

LLMs are terrible at asking questions. They just make a bunch of assumptions and brute-force something based on those guesses. Good human developers always pause to ask ...

  1. Mike Judge in his post "Where's the Shovelware? Why AI Coding Claims Don't Add Up" https://mikelovesrobots.substack.com/p/wheres-the-shovelware-why-ai-coding (discovered this link via Dan's email, thank you, Dan!)

Was I hoodwinked by the screens of code flying by and had no way of quantifying whether all that reading and reviewing of code actually took more time in the first place than just doing the thing myself?

If so many developers are so extraordinarily productive using these tools, where is the flood of shovelware?

  1. SingularitySkyNet's post in Cursor forums https://forum.cursor.com/t/why-the-push-for-agentic-when-models-can-barely-follow-a-single-simple-instruction/137154 and the HN discussion https://news.ycombinator.com/item?id=45577080

Why the push for Agentic when models can barely follow a single simple instruction?

  1. In the HN thread for Simon Willison’s blog titled "Vibe Engineering" https://news.ycombinator.com/item?id=45503867:

simonw: I think existing software development skills get a whole lot more valuable with the addition of coding agents

lemming: While this is true, I definitely find that the style of the work changes a lot. It becomes much more managerial, and less technical. I feel much more like a mix of project and people manager, but without the people. I feel like the jury is still out on whether I’m overall more productive, but I do feel like I have less fun.

  1. Prahlad Yeri's "I am a programmer, not a rubber-stamp that approves Copilot generated code" https://prahladyeri.github.io/blog/2025/10/i-am-a-programmer.html and the HN discussion https://news.ycombinator.com/item?id=45588283

  2. An Ask HN, "Has AI stolen the satisfaction from programming?" https://news.ycombinator.com/item?id=45572130

  3. A 15 minute talk by CJ, from the YouTube channel Syntax in the video "AI Coding Sucks" https://www.youtube.com/watch?v=0ZUkQF6boNg

I used to enjoy programming. Now, my days are typically spent going back and forth with an LLM, and pretty often yelling at it, or telling it that it is doing the wrong thing and getting mad that it didn't do what I asked it to do to begin with.

It's not fun. It's not fun at all. Why did I become a programmer? One of the reasons I like programming is that it is predictable, logical, knowable. You can look into the documentation, under the hood, into the source code, decompile something or watch network traffic... once you have an idea of how something works, you can be sure it's going to work the same way the next time you look at it.

Computers are logical systems. Programming languages are formal logical languages, and that works really well with my brain. When we are working with AI and LLMs it is not predictable. You can use the exact same prompt and get a different result every single time.

Honestly, that's not what I signed up for.