Pragmatic dotNET

From Vibe Coding to AI Workflow

How to turn AI-assisted coding into a repeatable process - from knowledge gathering and specification through implementation and review to keeping AI use under control. Based on real project experience.

Why a Single Prompt Is Not Enough

From Vibe Coding to AI Workflow: Why a Single Prompt Is Not Enough

AI workflowLLMSoftware deliveryVibe codingSDLC
AI can quickly make software development look easier than it really is. You type a prompt, get code, and sometimes even a complete working project. With a small project or an isolated part of a system, everything looks good. This becomes a problem when that code has to fit into an existing system that already has its own rules, past decisions, and constraints - not an empty repository.
What Controls an Agent?

From Vibe Coding to AI Workflow: What Controls an Agent?

AI workflowLLMSoftware deliveryAI agentsSDLC
In the first post, I explained why a single prompt is not enough in a larger project. The natural next step is an agent, but that word gets used very broadly. Before talking about tools and configuration, we need to define what we actually mean by an agent in AI-assisted software work.