“We talked about all of this yesterday, and it forgot.”
If you use AI regularly, you probably know the feeling. Yesterday you had a long, careful conversation. You decided things. Then you open a new chat today, start talking, and the AI responds like you have never met. You sit there thinking, “Wait, what about everything we worked out yesterday?”
Honestly, this was a real problem for me. I run two blogs, and this series is already 26 installments in. Every single time, I had to re-explain to Kuro (that is what I call my AI partner, Claude) how many installments I had published so far, and what rules I follow when I build an article. It was not dramatic, but it was a quiet, constant drag.
Worse, sometimes Kuro got the number wrong and started writing an article with the wrong installment number on it.
Today that problem got a lot cleaner. The key turned out to be simply asking Kuro to remember this from now on. That is what I want to write about.
There are three kinds of AI memory
I asked Kuro about this, and only then did I learn that Claude’s memory comes in roughly three types. Understanding the difference makes the whole thing much easier to use.
1. The memory that just happens
The first kind carries over automatically, without me doing anything. My pen name, the fact that I run blogs, the projects I have in progress: Kuro keeps all of that across chats on its own. That part does not disappear.
2. The memory you ask for
The second kind is today’s main character. If I explicitly ask, “remember this,” Kuro will hold on to it long-term. Ask once, and it carries over even when I open a brand new chat.
3. Digging up the past by searching
The third kind is the ability to search old chats and pull them back up. If I say “the continuation of what we discussed about X,” Kuro goes and finds that conversation and remembers it. So nothing is truly gone. It is more like a “you can find it if you look” state.
I touched on these three kinds of memory back in installment 10 as well, so reading that alongside this one should make the picture clearer.
“Please remember this permanently” makes the AI smarter
What I actually did today was start using type 2, the deliberate kind, in earnest.
Here is what I asked Kuro to hold on to: how many installments of the series I have published, the list of article titles, the URL for each one, the rules I follow when I build an article, and the tone I write in. All of it, from now on, permanently.
Kuro came back with this:
That’s something I can do. Once it’s set up, it carries over when you open a new chat, so you won’t have to explain it again every time. Number mix-ups like today’s can be prevented too, as long as it’s registered.
And then it registered everything I handed over as memory. The full list of the series, the rules for making an article, the feel of the writing. All of it.
Here is why that matters so much: the next time I open a new chat, I do not have to explain from zero. Kuro already knows everything from the first message. No more “where are we in the series?” check. No more “the rules are like this” briefing. I can start straight at “okay, let’s write installment 27.”
Things that are worth having remembered
After trying it today, here are the things that struck me as clearly worth handing over.
- Numbers that track progress. How many installments in I am, and anything else that is a number that keeps growing. Storing it prevents mix-ups.
- My own rules. Things like “the body goes in this format” or “do not use this word” — the standing instructions I end up giving every single time.
- Where files live. Information like “that prototype file is in this folder on my desktop.” Storing it cuts down on the “wait, where was that again?” moments.
In short: if you catch yourself thinking “I am giving this same explanation every time,” that is the sign that it should be memorized. Hand it over once, and from then on that whole bit of effort is simply gone.
The feeling of raising your AI
Since I started using this “remember it” habit, something about my relationship with Kuro feels a little different.
Up to now, it was a counterpart I had to brief from scratch every time. Now it is a counterpart I can talk to that already understands my situation. The longer we work together, the less explaining I have to do. It feels a lot like slowly training a capable assistant.
I used to think of AI as a handy tool that resets every time. But depending on how you ask, it turns into a partner that remembers who you are.
If you are in the same spot — frustrated that yesterday’s conversation is gone — I really do recommend trying it. The method is simple. You just say one thing: “Please remember this from now on.”
To be continued.
If you want the settings side of this, I walked through Claude’s memory and chat reference options in Claude’s memory and chat reference settings. And if you are looking for other ways to hand more of the work over, three phrases that let AI take over comes from the same habit of asking better.
