Siddharth Saoji The more I use LLMs for learning, brainstorming, and decision-making, the more I notice something subtle — and a little unsettling.
When you stick with one tool like Claude or ChatGPT over time, it accumulates memory about you. It starts referencing things only a friend would know. And that familiarity is quietly powerful — it lowers your skepticism. The answers feel more relevant, more personal, and so you start trusting them more.
That’s exactly where it gets tricky.
These models will never have the full picture. They don’t know how your specific situation is evolving, how the people around you will react, or how fast the world has shifted since their training cutoff. In a space like AI, where the landscape changes week to week, that gap matters. And there’s another layer: LLMs are trained on what people write, not necessarily what people think or feel.
Those are different things — especially when you’re making grey-area decisions about a career move, a business bet, or where to buy a house.
Most real decisions aren’t black and white. They’re contextual, emotional, and deeply human. An LLM can give you a direction, but it won’t always know the consequences of you taking it.
I don’t have a clean fix for this. But it’s worth remembering: the more familiar a tool feels, the more critical you need to stay.