tiny-essentials
Version:
Collection of small, essential scripts designed to be used across various projects. These simple utilities are crafted for speed, ease of use, and versatility.
52 lines (33 loc) • 3.98 kB
Markdown
## 🧭 How I Work with AI (Tips & Workflow)
Over time, I've built a personal workflow to make sure AI helps me *efficiently* — without ruining my focus, logic, or the structure of my code and documentation. Here are some habits and tricks I always follow:
* 📝 **Persistent Prompt Instructions**
I noticed that AI often starts forgetting layout and formatting instructions after just a few messages. To prevent this, I keep a notepad with my core prompts and manually repeat key instructions every time I ask something, especially when working on documentation.
* ⚙️ **System Prompt Setup**
One of my most effective strategies is maintaining a *System Instruction prompt* that defines the main behaviors I expect in **all my projects**. It includes things like:
* Avoid adding humor or randomness
* Always be 100% sincere and logical
* Don’t include jsDoc descriptions unless I explicitly allow it (especially when I'm still working on a WIP feature)
* 🧠 **How I Talk to AI**
I don’t try to talk to AI like it’s a person. I speak in a very natural, clear, and structured way — as if I were talking to a computer, because that’s what it is. This keeps things predictable and productive.
* 🔧 **Using the Chat History Interface Carefully**
I frequently use history editing tools (when available) to fix incorrect messages and keep the whole conversation aligned and accurate. The cleaner the context, the better the answers — especially over long sessions.
* 🧹 **Avoiding Context Drift**
The more you keep talking without resetting or correcting the prompt, the more likely it is that AI starts misunderstanding the conversation. So I always try to keep things clean and well-structured to preserve context.
* 📏 **Limit Code Size per Message**
I avoid dumping extremely long blocks of code into the chat. Instead, I share only the relevant part or break the code into smaller, medium-sized chunks. This prevents confusion and makes it easier for AI to focus on what I actually need.
* 💬 **Managing Large Output Responses**
If a reply needs to be very long, AI might break or compress the message too much to fit it into a single reply — and this often ruins the meaning or structure of the output. I always consider this when asking questions and sometimes break complex requests into smaller steps.
---
## 🤖 About AI Usage in the Projects
Some parts of this project were assisted by AI, but all the core decisions and logic come directly from me. Here's how I usually balance things:
* 🧠 **Ideas and Logic**:
All the concepts, how the system works, and the reasoning behind each part of the code are my own. I use AI mostly as a helper to speed things up or explore alternatives, but I’m always the one guiding and correcting everything.
* 📄 **jsDocs and Type Structures**:
I take care of the structure and typing in the documentation myself. When it comes to writing the descriptions, I might use AI to help word things better, but the meaning and context are always mine.
* 🧪 **HTML Test Files**:
For testing interfaces, I usually let AI generate the base since it's faster to describe what I need than build it all from scratch. It helps me quickly test the behavior of what I’ve programmed.
* ✍️ **Text (Messages, Descriptions, etc.)**:
I often ask AI to help me write or polish messages, explanations, or descriptions. It's great for keeping things clear and readable, but I always review and adjust it to match the logic and personality I want.
* 🌸 **Style and Personality**:
I love using emojis and I’ve set up my system to guide how AI writes with my personal style. That includes tone, formatting, and the little details that make the project feel like mine.
This approach is consistent across all my projects. It lets me stay focused on what really matters to me — logic, structure, and creativity — while still using AI as a practical tool when it makes sense.