The Problem Everyone Tells You About
No setup in this lesson. Nothing to install. Just the thing you have to understand before any of the rest makes sense.
It forgets you completely
Tell Claude about your dog today. Open a new conversation tomorrow and it has no idea you have a dog.
It isn't being rude, and it isn't a bug someone forgot to fix. Here is what actually happens each time you send a message:
- The AI reads your message
- It writes a reply
- It keeps nothing
Next time, it starts over from zero.
Why it works that way
Picture someone who is brilliant at their job and has no memory at all past the end of the current conversation.
You can hand them a stack of paper. They will read every page, understand it completely, and give you a genuinely good answer. Then they hand the stack back, and everything on it is gone from their head.
That's the model. It is very good at working with what's in front of it right now, and it retains nothing between one conversation and the next.
So when a chat app "remembers" your name, something outside the AI wrote that down and quietly slipped it back into the stack of paper before your next question. The AI didn't remember. Something handed it a note.
That's the whole trick, and it's what you're going to build — properly.
The stack has a limit
One more thing to know, because it explains a lot of odd behaviour.
The stack of paper has a maximum height. You'll hear it called the context window. Everything the AI can consider — your question, the conversation so far, any notes handed to it — has to fit.
When a long conversation starts going in circles or forgets what you agreed twenty minutes ago, that's usually the bottom of the stack falling off the desk to make room at the top.
This is why "just paste in everything you know" is not a plan. There's a ceiling, you'll hit it, and past that point adding more pushes something else out.
So the fix is obvious, right?
Save the conversations. Look them up later. Paste the useful bits back in before the AI answers.
That's it. That's what nearly every "give your AI a memory" tutorial teaches, and it is genuinely the right starting point. You'll build exactly that in Course 04.
It works. It works well.
For a while.
Try it yourself
Two minutes, and it makes the next lesson land properly.
1. Open claude.ai or chatgpt.com and tell it something specific about yourself. Not "I like dogs" — something with a detail in it. "My dog Ranger is a nine-year-old shepherd mix and he's on a prescription food."
2. Now open a brand new conversation. Not a new message — a new chat.
3. Ask: "What food is my dog on?"
Watch it not know.
4. Now, in that same new chat, paste your original sentence first, then ask again.
It answers perfectly.
That gap — between step 3 and step 4 — is the entire problem. And step 4 is the entire fix, done by hand.
Everything from here is about doing step 4 automatically, for thousands of notes, without you having to remember what to paste.
What's next
The next lesson is the one almost nobody teaches, and it's the reason this course exists.
The obvious fix breaks. Not immediately — around month four, quietly, in a way that looks like the AI just got worse.