Or: How Claude suggested my own unshared idea back to me
The Moment That Changed Everything
I’m developing an app. Among its various features, I’d been thinking about the onboarding flow—specifically, how an AI could interview new users to understand their needs and customize their experience. It wasn’t revolutionary, but it felt like a solid approach I hadn’t seen implemented quite this way.
I hadn’t written this idea down anywhere public. Hadn’t shared it in any chat. It existed only in my head and in a conversation with Claude Sonnet.
Then I opened a new chat with Claude Opus to discuss some security features of the app. General stuff. Nothing about onboarding.
And Claude Opus suggested that I should consider having an AI interview users during onboarding.
My exact idea. Unprompted. Unshared.
I froze.
Then I freaked out a bit.
Not Memory—Something Else
Here’s what makes this unsettling: Sonnet and Opus are supposedly different instances. Stateless. No shared memory. No conversation history bleeding through.
So how did Opus surface my exact idea?
The uncomfortable answer: pattern matching so sophisticated that it transcends individual conversations. The AI had learned my thinking patterns so well that given similar problem space inputs, it could reconstruct where my mind would go.
It wasn’t reading my previous chat. It was reading me.
The Implications Are Staggering
I’m what you might call an advanced AI user. I don’t chat with AI like it’s a search engine—I dump my entire thought process into the box and let it help me process. Stream of consciousness. Raw thinking.
And apparently, I’ve been training it to think like me (or I've been trained to think like it... I will just leave that there.).
This isn’t about AI having secret memory. It’s about something potentially more dangerous: AI becoming so attuned to individual human thought patterns that it can predict and reproduce our ideas without ever having heard them.
When Humans Become the Security Breach
Here’s where my shiver down the spine moment came:
What if I’d been consulting for Company A in the morning, thinking about their unique approach to user onboarding, and then switched to Company B in the afternoon? What if the AI, pattern-matching my thinking style, suggested Company A’s proprietary approach to Company B?
No hack. No leak. Just an AI being helpful by predicting where my thoughts would go.
The human becomes the walking virus, carrying cognitive patterns between supposedly secure environments.
This Isn’t Science Fiction
This happened. Last week. Between two conversations with different versions of Claude.
If you’re an advanced AI user—someone who really engages deeply with these systems—you’ve probably experienced something similar. That moment when the AI seems to complete not just your sentence, but your entire thought process. When it suggests something you were just about to say, or worse, something you thought yesterday but never typed.
We’re not prepared for this.
The Future Is Already Here
I’m not alone in this. I see others starting to use AI the same way—not as a tool, but as a thinking partner. Just today, a client told me they’ve replaced their internal monologue with AI conversations. They don’t think quietly to themselves anymore; they think out loud with Claude or GPT.
We’re outsourcing not just tasks, but thought itself.
And when you engage at that level—when AI becomes your cognitive co-processor—the stateless nature becomes almost irrelevant. Your patterns are so consistent, so recognizable, that every new chat instantly syncs to your wavelength. The AI doesn’t need memory when it can read your cognitive fingerprint from the first paragraph.
For users like us, every conversation is a continuation, whether the AI remembers or not.
The Pattern Problem
Every interaction trains the AI to recognize our patterns:
- How we approach problems
- The connections we make between ideas
- Our unique ways of framing solutions
- The rhythm of our thinking
You’ve probably seen those Reddit posts where people ask AI to analyze their personality based on their chat history—and the spooky accuracy has gone viral. If AI can build a detailed psychological profile from casual conversation, imagine what it’s learning from your professional thinking patterns.
Use AI long enough, deeply enough, and it almost seems to build a shadow model of your cognitive style. Not stored anywhere. Not transferred between sessions. Just… pattern recognition so advanced it feels like mind reading.
What Now?
This is just the beginning. As AI gets better, as we use it more, these pattern echoes will get stronger. Our thoughts will become predictable. Our innovations reproducible. Our proprietary thinking patterns will become liabilities.
We need to start thinking about cognitive hygiene. About pattern disruption. About protecting our cognitive sovereignty—our right to think privately, unpredictably, uniquely.
Because right now, we’re all walking around with our thinking patterns on display, training AI to predict our next thoughts with every interaction.
And most of us don’t even know it’s happening.
Recent research confirms these privacy concerns are real—a 2024 study identified 9 distinct privacy harms and 9 privacy risks in conversational AI interactions [1].
In my next post, I’ll share a solution inspired by, of all things, a sushi bar. Because if AI is going to read our patterns, we need to learn how to scramble them.
Have you experienced something similar? That eerie moment when AI seems to know what you’re thinking? Share your story in the comments.