Most people use AI assistants the wrong way. They open a chat, paste a generic prompt, and then complain when the code comes back looking like it was written by a junior developer who read a textbook last week.
If you want the AI to write code that matches your style, matches your architecture, and follows your naming conventions, you have to feed it your personal rules. You have to train it to operate the way you do.
Here is how you can build a personal configuration file that forces the AI to think and write like you.
The Problem with Default Settings
Out of the box, AI models are trained to be polite, wordy, and generic. They want to please you, so they default to the most common patterns they found on GitHub.
If you write code for a living, you probably have habits. Maybe you hate em dashes in your comments. Maybe you prefer tailwind utility classes grouped in a specific way. Or maybe you have strict performance rules about how database queries are structured.
If you do not tell the AI these rules, it will guess. And its guess will be a generic, average representation of the internet.
To fix this, you need a single source of truth: a file that defines your professional identity and coding standards.
Step 1: The Interview Prompt
Instead of writing this file from scratch, you can have the AI build it for you by interviewing you.
Start a fresh chat with your favorite LLM and paste this exact prompt:
I need you to generate an about-me.md file.
This file defines who I am(role, company, target audience), How I think, How I write, how I work (tools, processes, what "done" looks like), what good looks like (standards, examples, what to avoid), what "wrong" looks like (bad patterns, what annoys me), my rules (hard lines, non-negotiable), instructions ( 10 rules: what to do and what not to do ). Basically how I operate so my AI assistant can think like me. Interview me using AskUserQuestion
The AI will ask you a series of questions. Answer them honestly. Mention the tools you use, the styling rules you care about, the bugs that annoy you, and what your definition of "done" is.
Step 2: Refining the Rules
Once the AI generates your file, refine it. Split your preferences into clear categories:
- Role and Context: Who you are and what platforms you build.
- Standards: What good code looks like in your project.
- Anti-Patterns: What patterns, imports, or stylistic habits the AI must avoid.
- Hard Rules: Non-negotiable instructions (such as "never use placeholders" or "always use absolute imports").
Keep the rules concrete. Instead of saying "write clean code," say "keep functions under 50 lines and write explicit error handlers."
Step 3: Feeding the Context
Once you have your profile file, you need to make sure the AI reads it. Depending on the tool you use, you have a few options:
- System Prompts: If you use a web interface, paste the contents into the custom instructions box.
- Rules Files: If you use editor extensions like Cursor or Claude Code, save the file as a system rule (like
.cursorrulesor.agents/rules/about-me.md) in your project root. - Context Pinning: Pin the file to your active chat sessions so the assistant references it on every prompt.
The Payoff
It takes about fifteen minutes to set this up, but it saves hours of refactoring later. When the AI has your profile pinned, it stops suggesting deprecated libraries. It stops writing corporate copy in your blog drafts. It stops adding lazy placeholder comments.
If you treat the AI like a generic search bar, you get generic results. Treat it like an apprentice that needs onboarding, and it will start coding exactly like you.