I've worked in IT for about five years and currently manage a department. I'm not a hands-on developer, but I'm comfortable around coding environments, technical jargon, software processes, and the ethical discussions that come with them. I also completed some beginner programming courses and small university projects several years ago, so I'm not starting from zero. I still have a virtual machine, VS Code, and a Git repository containing those old projects, although I haven't coded independently in a long time.
I'd like to use AI—probably Claude—to build a few personal projects, such as a weather report, a personal weight-loss tracker, and eventually a tool for comparing statistics or items in a game. I might also make very minor UI improvements to our SaaS product if the opportunity comes up, though I understand the risks of using AI-generated code at work and that any changes would need proper review and approval.
Rather than simply asking an AI to "make me a website," I want to establish good foundations first. I'd like recommendations for clear videos or guides covering project setup, choosing an appropriate language and tools, version control, testing, documentation, and sensible AI-assisted development practices. My goal is to understand and review the code, ask useful questions, and minimize low-quality AI-generated code or bad habits. What would be a good way to approach this?
2 Answers
A good AI-assisted workflow is to divide the project into small pieces and handle one meaningful task at a time. Don’t ask for the whole application in one prompt. Start a fresh session for a major feature when the existing context becomes cluttered, and correct or revise an earlier instruction when the AI has misunderstood rather than piling on more contradictory prompts.
For Claude-based projects, learn how project instructions work, especially CLAUDE.md files, reusable skills, and any project rules supported by your setup. You can keep general conventions in a main instruction file and add more specific files in subdirectories. Include things like coding style, file organization, testing commands, architectural decisions, and assumptions. It’s also useful to ask the AI to create a plan first in a read-only or planning mode, identify assumptions for you to confirm, and maintain concise technical notes that can be loaded only when relevant.
The exact token-optimization tools are optional. Good project structure, clear instructions, small tasks, tests, and human review matter more than squeezing every possible token out of the workflow.
You probably don’t need a long AI-specific tutorial before starting. Begin by choosing the right language and project type, either with help from an experienced developer or by asking the AI to compare a few reasonable options. Then use an IDE such as VS Code with suitable language support and an AI extension.
The important part is to avoid treating AI as a replacement for understanding. Give it a specific, manageable task rather than the entire project, inspect what it produces, run tests, and ask questions about anything you don’t understand. This can be a useful way to rebuild your programming skills while working on something practical. Any code used professionally should still be reviewed and validated by developers.
That makes sense. I already have VS Code installed from my university projects, so I’ll review the setup and probably start with a clean project rather than immediately asking for a complete application. At work, we already require people to take responsibility for and review any AI-assisted code, so I’m familiar with that limitation.
This is the kind of practical checklist I was looking for. I understand the general idea of splitting work into smaller pieces, but the guidance around project instruction files, planning first, and recording assumptions gives me a much better starting point.