PN
GitHubLinkedInX
Banner
Back to portfolio
June 27, 2025 • 3 min read

LLMs Are Getting Memories. What Happens When AI Starts Remembering You?

We've been obsessed with making AI smarter. More context. Faster answers. But now, something new is happening.

AI is learning to remember you.

Not just in one session.

Across days and even weeks and it changes everything.

So, What even is LLM Memory?

Up until recently, Large language models like ChatGPT, Gemini, Claude were like goldfish.

Every time you started a new chat, it was a fresh state. You could tell GPT your name, your tech stack, or your idea of the project or so and next time? Gone.

But now?

LLMs are developing memory.

This means :

  • They remember facts about you (like your name, or even your members)
  • The keep context across multiple conversations.
  • They even remember your mistakes and help you avoid them again in the future 💀

    • Why Is This Such a Big Deal?

    Because now… AI is becoming personalized.

    We're entering a future where:

  • Your AI doesn't just help - it knows you.
  • AI tools become actual teammates and not just a one-off tools.

    • What Tools Are Doing This Already?

    Let's take a look at the LLMs currently flexing their memory muscles:

    ChatGPT

    Memory is now on by default for ChatGPT Pro.

    You can see what it remembers and edit/delete it.

    Example: It remembered I was working on a solo travel app and suggested API integrations without me prompting it again.

    Claude 3.5 Sonnet

    While it does not have a long term memory yet, it is insanely good at recalling things within large contexts (like the whole codebase that is submitted to it)

    When memory launches, it'll likely be crazy strong due to Anthropic's focus on alignment and safety.

    So… Is This Good or Creepy?

    Let's break it down.

    Pros:

  • Acts like your second brain
  • Feels like a real collaborator, not a tool
  • Gives personalized responses

    • Cons:

  • Privacy risks—what if someone hacks your AI memory?
  • What if it remembers things you regret saying?

    • The Privacy Angle: Are You Safe?

    OpenAI allows users to:

  • View/edit/delete memory
  • Turn memory off entirely

    • But here's the real thing: in the future, when every app has "memory", data ownership will matter more than anything.

    TL;DR

  • LLMs like ChatGPT now have memory
  • That means they remember you across chats
  • It makes AI feel more human-but also raises big questions
    • "We used to train AI. Now it's training itself-on us."