7 posts tagged with "vibe-coding"

AI coding assistants can be incredibly helpful — until they aren't.

Recently, I asked my AI assistant to help rebase my local main branch onto an upstream repository. What followed was a seemingly helpful but quietly destructive sequence of Git operations, carried out without sufficient warning or consent. The AI confidently moved forward, using commands like git reset and git push --force-with-lease — powerful tools that should never be used lightly.

Hook​

I've known for a while that vibe coding tools have a knack for cluttering codebases with duplicate files. That's exactly why I kept a close eye on my latest session with Cursor paired with Claude-3.7-sonnet. But even with my vigilance, these AI tools still managed to surprise me—by completely ignoring my explicit instructions to clean up duplicates. Instead of tidying my repository, they left a handful of scripts scattered across directories like forgotten party favors.

"Hey, could you clean up those duplicate files?" I prompted clearly, expecting thoroughness. What I got instead was a comical, frustrating hunt through leftover scripts, turning a routine cleanup into an unexpected treasure hunt.

When Your AI Feels Like a Junior Dev: The Growing Pains of Vibe Coding Tools​

Introduction​

Have you jumped on the "vibe coding" bandwagon yet? If you haven't heard, vibe coding is all about coding alongside AI assistants like Cursor and Claude. These tools promise to make your coding smoother, faster, and maybe even a bit more fun. Sounds great, right?

Well, mostly. The reality is, sometimes my AI sidekick feels less like an expert coder and more like that enthusiastic junior dev fresh out of bootcamp. They're eager, helpful, and… prone to rookie mistakes.

Let me dive into some entertaining real-world examples showing how my vibe coding tools sometimes patch symptoms instead of truly solving problems.

I've been experimenting with "vibe coding" tools like Cursor and Claude Code (3.5)—the kind of AI pair programmers that promise to understand your code, suggest fixes, and help you fly through bugs without touching the keyboard too much. In theory, it's magical.

In practice? Well... let me tell you a story.

Context​

Recently, I dove into creating a blog site dedicated to my new passion: vibe coding. I've previously built projects using Docusaurus, a React-based framework optimized for documentation and blogs, making it a natural choice for this project. Initially, I considered using Hashnode for hosting based on Deepseek's recommendation. At first glance, Hashnode appeared to be a great fit—it's easy to set up, visually appealing, and community-oriented. However, a deeper dive revealed serious SEO issues, specifically broken sitemaps and landing pages that couldn't be indexed by Google. Given the importance of SEO for visibility, this was a significant drawback. With that, I quickly decided to migrate my content from Hashnode to a more reliable solution—Docusaurus.

Embarking on the journey of translating C++ code into TypeScript, I enlisted the assistance of Claude 3.5 Sonnet, Anthropic's AI language model. My objective was to achieve a faithful, line-by-line translation of Config.hpp and Config.cpp, ensuring type safety by utilizing TypeScript types that closely mirror the original C++ structs and classes. The aim was to preserve the original file and class structure.

Ever been deep into a vibe coding session, grooving along with your AI assistant, only to have your flow completely disrupted because your AI buddy stubbornly refuses to admit it's made a mistake? Yep, same. Today, let's dive into the quirky world of vibe coding and unravel one of its biggest headaches—LLMs' pesky self-reflection limitations.