Vibe coding is a good starting point, but it is not where serious AI-assisted development ends. The next step is agentic engineering: using AI coding agents inside a controlled engineering workflow, with context, tests, review and clear boundaries. Vibe coding often focuses on the generated output: Ask for feature -> get code -> run it -> ask for fixes Agentic engineering focuses on the system ar
More rules should mean better output. That's the intuition. I spent weeks building a comprehensive CLAUDE.md — 200 lines covering naming conventions, security rules, error handling, architectural patterns, import ordering, type safety requirements, and more. I was proud of it. I'd thought through every scenario. Then I scored the output. 79.0 / 100. My carefully crafted documentation was actively
This post was created with AI assistance and reviewed for accuracy before publishing. Cursor can use project rules and documentation to steer behavior. Exact file names and mechanisms evolve; check Cursor documentation for the current layout (for example rules in .cursor or legacy .cursorrules patterns). Short, enforceable bullets beat long essays: stack versions, test commands, “no new dependenci
Vibe coding is one of those terms that sounds unserious until you notice how many people are actually doing it. The basic idea is simple: describe what you want, let an AI coding tool generate the implementation, run it, adjust the prompt, and keep going. It can feel magical. It can also go wrong very quickly. Vibe coding works best when the problem is visible and forgiving: small prototypes inter
Have you ever looked at code you wrote six months ago and thought: "Who wrote this monster?"? Relax, it happens to all of us. In software engineering, writing code that a machine understands is the easy part. The real challenge is writing code that other humans (including your future self) can understand, maintain, and scale. This is exactly where Software Design Principles come into play. In this
Part 1 of 5 in The New Engineering Contract — what it means to lead engineers when AI is doing more of the coding. SWE-CI tested 18 AI models across 71 consecutive commits. Most broke something on commit 47 they'd already broken on commit 1. That's not an intelligence problem. That's a learning system that isn't learning. A paper made me uncomfortable this month. Not because of what it found about