Hey! Ankur here, and this is the 17th edition of Lazy AI — 5 mins of reading to help you stay ahead of the AI curve.

Today’s issue might make you a little uncomfortable..but in a good way.

Before I begin, I’d really want to understand your background, so that I can write future editions keeping that in mind. So..please tell me!

A friend of mine — marketer, similar experience level — asked ChatGPT to write a cold outreach email for a SaaS product. Generic prompt, nothing fancy. The output was...just fine.

Usable, but also template-y.

I typed the exact same prompt into ChatGPT. But what I got back was a cold email that referenced B2B SaaS specifically, matched my usual tone, and even structured the email the way I like — short paragraphs, no fluff, CTA at the end.

It felt like it knew me.

Because it does.

And here’s the thing — I never sat down and told it all of this. Not explicitly, anyway.

So what’s going on?

Most people think of AI chatbots as blank slates. You type, it responds, you close the tab, it forgets everything. That was true once. It’s not true anymore.

Every major AI assistant — ChatGPT, Claude, Gemini — now has a memory system. And it’s not one system. It’s two.

That second layer is why my cold email was better than my friend’s. Not because I’m a better prompter. Because ChatGPT had been reading our past conversations and building a profile of how I work and how I like things written.

You don’t need to be an engineer to talk AI. Subscribe to this newsletter — every other day, you’ll understand AI well enough to be the smartest person in the room at work.

Each issue is just 5 minutes — less than the time you spend doomscrolling before bed. Except, this actually moves your career forward. Join 8,000+ subscribers now.

Here’s where this gets personal.

Last week, I got curious and asked Claude to show me everything it has stored about me. The explicit stuff — the saved memories I supposedly control.

I expected maybe 2 or 3 entries. It showed me 14 separate memory entries.

My LinkedIn strategy. My job details. My posting schedule. My career history. My content frameworks. Things I’d mentioned in passing during working conversations — not things I’d ever said “hey, remember this.” The AI had decided, on its own, that these facts were important enough to store permanently.

And that’s just the saved memories — the ones I can see and delete.

The contextual memory is a whole separate summary generated by a background system that processes my conversations every 24 hours and builds a running profile. I can’t edit it line by line. I can only see the output.

(I’m not going to lie — it was a weird feeling)

Let me walk you through what happens every time you open a new chat. Because it’s not “blank screen, fresh start” anymore.

All of this sits inside something called the context window.

For most AI chat apps, the desk fits roughly 200,000 tokens, which is about 150,000 words (up to 1 million tokens for developers using the API). Sounds like a lot, but memories, instructions, and conversation history eat into it with every message.

So when you type “write me a cold email,” the AI isn’t just reading those six words. It’s reading those six words plus everything it already knows about you.

Your AI isn’t smarter for one person. It’s more informed.

Here’s where it gets interesting — each AI company has drawn very different boundaries around this.

Claude is the most conservative — saved memories plus a daily auto-summary. No ecosystem integration. ChatGPT adds a chat history reference layer that pulls context from past conversations even when you didn’t ask it to save anything. Gemini goes the furthest — Google launched “Personal Intelligence” in January 2026 for paid US subscribers, letting it pull from your Gmail, Photos, YouTube, and Search history. When you ask for restaurant recommendations, it’s checking what you’ve searched, emailed about, and watched.

Same feature. Very different boundaries.

You don’t need to be an engineer to talk AI. Subscribe to this newsletter — every other day, you’ll understand AI well enough to be the smartest person in the room at work.

Each issue is just 5 minutes — less than the time you spend doomscrolling before bed. Except, this actually moves your career forward. Join 8,000+ subscribers now.

I’m not against AI memory. It actually makes the tool dramatically more useful. When Claude already knows my writing tone and work context, I save 20 minutes of setup on every conversation.

But most people have no idea this is happening. They think they’re talking to a blank slate. They’re not. They’re talking to a system that has been quietly building a profile of who they are, how they work, and what they’re likely to ask next.

I’m not saying this is evil. I’m saying you should know.

Because once you know, you can make it work for you:

Next time your colleague says “ChatGPT gives me amazing answers” and yours feel blah, it might not be the prompt. It might be what the AI already knows about them.

The person who’s been using the tool daily for six months, telling it their preferences, correcting its mistakes — their AI is working with context. Yours might still be guessing.

AI memory is a multiplier. But only if you know it’s there.

If this got you thinking, please share it with someone who keeps complaining that “AI doesn’t work for me” — they might just need to understand what’s happening behind the screen.

See you next time..

Cheers,

Ankur