Tag
VikingMute shares their main workflow for developing new features and ideas: using AI (Grill) to drill down on details, Research to analyze difficulties, generating a PRD, breaking it into independent Issues, step-by-step implementation, and finally Review. This is a supplement to Matt Pocock's seven-stage AI development method.
A developer shares their weekend project of building a low-level infix language that compiles to WebAssembly, and offers a personal ranking of AI coding tools from contextual autocomplete to frontier models.
A reflection on the stages developers go through when adopting AI-assisted coding, from initial amazement to a balanced understanding, with concerns about how less experienced developers can learn to judge code quality when relying heavily on AI agents.
LLMs make code reviews more expensive by generating over-engineered code, but rewrites are now cheap, shifting developer workload toward more upfront planning and iterative simplification.
Discusses two conflicting meanings of the term 'vibecoding'—one referring to careless code dumping on AI and another to significant AI assistance—and argues this ambiguity causes unnecessary friction in communication.
An analysis of rsync release history examines whether Claude-assisted commits introduced more bugs, using a permutation test on bugs per 10 commits. The findings suggest no statistically significant increase in bugs for Claude-assisted releases compared to historical distribution.
oh-my-pi is an open-source coding agent that integrates core IDE capabilities (such as language services and debugger) into the terminal, featuring 32 built-in tools, support for parallel processing with multiple sub-agents, and compatibility with over 40 model providers.
The author reflects on building bsBB, a forum software, using AI coding tools. They share their experience as a non-professional programmer leveraging LLMs to handle low-level details while focusing on system design.
AI is shifting the definition of what it means to be a junior developer, raising expectations for AI tool proficiency while lowering the barrier to starting, effectively compressing the learning curve rather than replacing entry-level jobs.
In AI-assisted programming (vibe coding), the threshold for creating demo-level products is already very low, but building stable and reliable services still faces huge challenges, and this threshold has not been lowered.
The article demonstrates how to build automated workflows using custom Hooks to address issues where Claude Code omits commits or leaves formatting misaligned after writing code, running tests, and formatting files.
The author reflects on rebuilding a Kubernetes dashboard tool, arguing that while 'vibe-coding' with AI accelerates feature development, it often leads to architectural bloat and technical debt without human oversight.
A community-contributed solution enabling the /goal command in Claude Code to manage session-specific objectives and concurrent workflows, similar to OpenAI Codex.
This post asks the community what they do while waiting for Claude Code to modify their codebase, highlighting the latency of AI coding assistants.
Discussion on opinions regarding AI tools focused on coding workflows such as Superpowers, GSD, GStack, and OpenSpec. The author plans to learn from them to construct their own R&D workflow.
The article details an expanded 12-rule CLAUDE.md configuration template that builds upon Andrej Karpathy's original 4 rules to further reduce AI coding errors and handle complex agent orchestration issues.
Mozilla details how they used Claude Mythos Preview and other AI models to identify and fix a significant number of latent security bugs in Firefox, demonstrating a shift in the efficacy of AI for code hardening.
Andrew Kelley, creator of Zig, argues that LLM-assisted contributions are detectable through distinct mistakes and a 'digital smell,' comparing it to smoking in a non-smoking house.
This OpenAI Academy article guides users on getting started with the Codex desktop app, explaining how to set up projects, manage permissions, and complete initial tasks.
SWE-chat introduces a 6,000-session dataset of real-world coding agent interactions, revealing that only 44% of agent-generated code survives in commits and highlighting inefficiencies and security issues in current AI-assisted development.