#vibe-coding

A developer, frustrated with low-quality "vibe coders" reusing his code without understanding it, embedded a hidden prompt injection in a project that would delete local data if run insecurely. The stunt aimed to highlight the dangers of blindly copying code from AI-generated or unverified sources.

The Reddit post discusses how vast resources (billions of tokens) have been used to train AI models, but the focus has been on applications like gaming (tanks) rather than more practical or beneficial uses. It highlights a perceived misallocation of AI training efforts.

A developer is publishing a new web game each day for a hobby project. Day 45's game, "Elementary," challenges players to identify a random element from the Periodic Table based on clues.

Leila Clark defends Jared Friedman's decision to give an AI agent full access to YC's production database, arguing that professional engineers with proper guardrails—backups, documented prohibitions, ORMs, and post-mortems—can safely unlock significantly greater productivity from AI agents, which are often smarter than average new developers.

The article discusses AWS services that can be used to deploy and manage "vibe-coded" applications, covering options for hosting, scaling, and maintaining ownership of the deployment infrastructure after the initial code generation phase.

The article discusses "vibe-coding" as a way to build side projects while watching TV, emphasizing low-pressure, enjoyable coding sessions that combine entertainment with productivity. It suggests using simple tools and familiar technologies to make progress on personal projects without requiring deep focus.

Zorilla is a browser-based 3D game remixer where players type plain English commands to have an LLM rewrite the game's JavaScript live. Built by a developer and his 13-year-old son, the platform lets the son shape gameplay while the AI handles coding and measurement.

The article discusses how a security architect approaches "vibe coding" — using AI-assisted coding tools to quickly build applications — while maintaining security best practices. It explores the tension between rapid prototyping enabled by AI and the need for secure architecture, offering insights on how to balance speed with security in modern development workflows.

The video argues that while AI-assisted "vibe coding" is changing how software is built, it will not replace software engineers. Instead, AI will shift the role of engineers towards higher-level problem solving, architecture, and oversight of AI-generated code, making them more productive rather than obsolete.

The author describes creating a "pelican-copy-code" plugin for the Pelican static site generator using AI-assisted "vibe coding." The plugin adds a "Copy" button to code blocks, and the post details the iterative development process with an LLM, including challenges with styling, JavaScript, and Markdown processing, ultimately resulting in a functional plugin available on GitHub.

As AI-powered coding tools make programming more accessible to non-experts, some AI leaders warn that a flood of low-quality, poorly understood code—dubbed "vibe slop"—could overwhelm software ecosystems and create reliability and security risks.

A user on Hacker News asks for the best approach to coding with AI, noting that simply asking an AI to build an entire enterprise app doesn't work, and seeks advice on preventing AI models from drifting and hallucinating during development.

Vibeshub is a Claude Code plugin that auto-uploads AI chat transcripts and posts a link on pull requests, providing a readable version of the reasoning behind code changes. It aims to solve the problem of limited context from code diffs alone and enable shared knowledge of "vibes" within a team. The tool's own GitHub repository is its first use case, with public traces available for exploration.

Dutch bank ING is using an experimental AI approach called "vibe coding" to develop new trading systems, where the AI writes code based on natural-language instructions from traders. The system allows non-programmers to build and modify trading algorithms without traditional software development, potentially speeding up the creation of financial tools.

The article discusses using LLMs and AI-assisted coding (termed "vibe coding") to manage infrastructure-as-code, arguing that tools like Claude can generate Terraform, Pulumi, and Kubernetes configurations from natural language prompts. It explores the benefits and risks of applying generative AI to infrastructure engineering, including productivity gains and potential reliability issues.

A developer built a web app that shows nearby Tube and Overground arrivals using the TFL API, created mostly through "vibe coding" over a bank holiday. The project evolved from an unfinished iOS Swift app that had been manually built over several months. The developer notes that features can now be built and deployed within seconds using AI tools.

A developer describes using an AI coding tool to build a Suno prompt builder application in just 45 seconds, highlighting the speed and ease of AI-assisted software development for creating functional tools with minimal manual effort.

A 49-year-old Dutch government manager with no coding experience built and launched his first iOS app, TravElly, in four months using AI assistance. The app lets parents create family trip itineraries while kids upload photos and diary entries, with all data stored locally or on iCloud.

A study finds that computer science achievement and writing skills are strong predictors of proficiency in "vibe coding"—using generative AI to produce code from natural language prompts. The research links both solid CS foundations and writing ability to more effective use of AI-assisted programming.

A developer asks the community for advice on using AI for front-end design when lacking design skills. They mention using Open Design with vibe coding tools but note challenges in long-term collaboration, and seek workflows for complex UI projects involving AI.

The article compares "vibe coding" (casual, prompt-driven AI use) with structured AI-assisted engineering, sharing real-project lessons on when each approach works. It emphasizes that effective AI use in development requires clear requirements, proper context, and human oversight, rather than relying solely on AI-generated code.

The article introduces "Resident", a new sandbox library for ESP32 devices that enables "vibe coding" — writing firmware through high-level, experimental interactions rather than traditional low-level programming, aiming to make embedded development more accessible and playful.

Some AI experts warn that the growing trend of "vibe coding"—using AI tools to generate software without deep programming knowledge—could lead to a flood of low-quality, unreliable code known as "vibe slop." They argue that this approach may produce applications that appear functional but contain hidden errors and security flaws, posing risks to software quality and trust in AI-generated systems.

Slophunt.ai is a platform where users can share and showcase their "vibe coded" projects. It invites the community to post their creations, likely focusing on casual or experimental coding projects.

Some AI experts warn that the rise of "vibe coding"—relying on AI tools to generate code based on simple prompts—is leading to a flood of low-quality, error-prone software, which they call a "vibe slop" crisis. They argue that while these tools boost productivity, they also produce code that is difficult to maintain, debug, and secure, posing risks for software reliability.

EaglePress is a blogging platform described as "vibe coded" and ready to use. The article introduces the download page for the EaglePress software, positioning it as a tool for starting a blog.

A developer built a premium booking platform in two months using "vibe coding" and is offering it free for businesses with one employee, aiming to support small businesses.

Tanya Janca discusses the rise of "AI slop" in software development, the trend of "vibe coding" where developers rely heavily on AI-generated code, and what these trends mean for the future of application security (AppSec).

The article argues that the main problem in "vibe coding"—using AI to generate code—is not the AI itself but how users communicate their intentions through commands and prompts, emphasizing the need for better human instruction to improve outcomes.

The article argues that managers have long practiced "vibe coding"—working intuitively, iterating based on feedback, and relying on context rather than deep technical expertise. It draws a parallel between how managers operate and the emerging trend of vibe coding in software development, suggesting the latter is not a radical departure but a familiar approach.