Working With an LLM
In the last session, we covered what a Large Language Model is and how it learns. It generates responses rather than looking them up, which means it can be wrong, and it does not remember you between sessions unless you give it a way to do so.
This session covers two things. First, a simple way to think about an LLM that will help you work with it well. Second, a tool called a .MD file, and why it solves that memory problem directly.
A Simple Way to Think About It
Here is an analogy that may help.
Imagine you hired a very well-read assistant. This assistant has read more books, articles, and documents than any human being alive. They are fast. They are articulate. They can write a draft in seconds. They can summarize a 50-page document in two minutes.
But this assistant has a few specific limitations you need to know about.
First, they do not always know when they are wrong. They may state something confidently that turns out to be incorrect, because they are drawing on patterns from their reading rather than verified facts.
Second, they do not remember previous conversations unless you give them notes. Every morning, they show up with no memory of what you discussed the day before.
Third, they only know what was in the material they read during their training. If something happened after that training ended, they may not know about it.
Fourth, the quality of their work depends heavily on the quality of your instructions. Vague instructions produce vague results. Specific instructions produce specific results.
If you understand those four things, you understand how to work with an LLM productively. You verify important facts. You provide context at the start of each session. You give clear, specific instructions. And you treat the output as a strong first draft rather than a final answer.
The Memory Problem, Revisited
We already established that an LLM does not remember you between sessions. Every time you open a new conversation, it starts fresh.
Think about what that means in practice.
Suppose you are using an AI tool to help you run a small business. You want it to help you write emails to customers. You want it to help you draft quotes. You want it to understand your products, your tone, and your typical customer.
If the AI starts fresh every session, you have two choices. You can re-explain all of that every single time, which wastes time and is frustrating. Or you can find a way to give it that background information quickly and consistently at the start of each session.
That second option is where .MD files come in.
What a .MD File Is
MD stands for Markdown. Markdown is a simple way of formatting plain text so that it is easy to read and easy for computers to process.
You do not need to be a programmer to use it. You do not need any special software, though there are free tools that make it easier. At its most basic, a Markdown file is a plain text file with a few simple symbols added to show structure.
Try It Yourself
Type or paste some text below using #, ##, ###, ** and *, and watch it turn into real formatting underneath. Clear out the example first and try your own.
Here is how it actually looks once rendered:
The pound signs create headings. Two pound signs create a smaller heading. The text underneath each heading is plain, readable content.
That is it. There is no code. There is no programming. It is just organized text with a few simple symbols.
When you save that text as a file with the extension .md, you have a Markdown file.
Why AI Tools Handle Markdown Well
AI tools, including the major ones like Claude, ChatGPT, and others, are very good at reading Markdown files. There are a few reasons for this.
First, a large portion of the text those tools were trained on was formatted in Markdown. Technical documentation, GitHub repositories, online tutorials, and many other sources use Markdown. So the AI has seen a lot of it and knows how to read it well.
Second, Markdown creates structure. Structure helps the AI understand what is important and how different pieces of information relate to each other. A heading tells the AI: this is a category. The text below the heading tells the AI: this is the content within that category.
Third, Markdown is lightweight. It does not add unnecessary formatting codes or hidden characters the way a Word document does. What you see is what is there, and the AI can process it cleanly.
For our purposes, that means a well-organized Markdown file is one of the most efficient ways to give an AI the background information it needs.
How to Format Your .MD File
| Symbol | What It Does |
|---|---|
| # | Biggest heading (title size) |
| ## | Medium heading |
| ### | Smaller heading |
| #### | Even smaller heading |
| **text** | Bold text |
| *text* | Italic text |
| ***text*** | Bold and italic together |
| ~~text~~ | |
| - text or * text | Bullet point |
| 1. text | Numbered list |
| - text (indented) | Sub-bullet, nested under another bullet |
| [text](link) | Clickable link |
|  | Picture, instead of a link |
| `text` | Code or a symbol shown as typed, not turned into formatting |
| ```text``` | Block of code, set apart from the rest of the page |
| > text | Quote block |
| --- | Line across the page (a divider) |
| - [ ] text | Empty checkbox |
| - [x] text | Checked checkbox |
| \ before a symbol | Shows the symbol as-is instead of using it for formatting |
| Two spaces at end of line | Starts a new line without starting a new paragraph |
| | text | text | | Table column |
Lets Prepare Some Files
Now we are at the end of lesson two and it is time for you to prepare two MD files for your AI. Below you will see two copy buttons. They will each copy the associated file to your computer’s clipboard. You can then paste them into a text editor. Move each file to an appropriate location and name it something you will remember similar to Your [FirstName_LastName_Personal_DNA.md]. Answer the questions to the best of your ability and save the files following the instructions below.”
Click Copy below, then paste the text into Notepad (Windows) or TextEdit (Mac). Fill in your answers, then save the file as:
Your FirstName_LastName_Personal_DNA.md
PERSONAL DNA Name: Location (city/state): Current Occupation: Household (spouse, kids, others I make decisions with or for): How I Like AI to Talk to Me Communication style I prefer (blunt and short, or patient and thorough): Future Direction Where I want to be heading: Goals (Top 5 for the next 90 days): 1. 2. 3. 4. 5. Why these goals matter to me (purpose/values/impact): What's On My Plate Right Now Current situations or decisions I'm working through: 1. 2. 3. Best Win & Hardest Lesson Biggest Achievement (what made it possible): Biggest Failure (what I learned, how I adapted): Constraints & Fears Time constraints, bottlenecks: Fears or risks I tend to avoid: 1. 2. 3. How I Expect to Use AI Recurring tasks or questions I'll bring to it: 1. 2. 3.
Click Copy below, then paste the text into Notepad (Windows) or TextEdit (Mac). Fill in your answers, then save the file as:
Your FirstName_LastName_Business_DNA.md
BUSINESS DNA Business Name: What I Sell or Offer: Who I Serve (my typical customer): How Long I've Been Doing This: My Tone & Voice How I want my business to sound (friendly and casual, expert and formal, blunt and no-nonsense): Words or phrases I use often when talking about my work: Words or phrases I never want used to describe my business: What Makes Me Different What I do better, faster, or differently than competitors: Why customers choose me over someone else: Goals for This Business Top 3 goals for the next 90 days: 1. 2. 3. Where I want this business to be in a year: What's On My Plate Right Now Current challenges or bottlenecks: 1. 2. 3. Decisions I'm currently working through: Best Win & Hardest Lesson Biggest business win so far (what made it possible): Biggest business setback (what I learned, how I adapted): Success Metrics What "a good month" looks like in numbers or outcomes: How I'll know this business is working the way I want it to: How I Expect to Use AI For This Business Recurring tasks or questions I'll bring to it: 1. 2. 3.