It's all about the memory...
Why your favorite AI might just be the one that knows you best.
Yesterday, while cross publishing my test of Claude Design, I received this:
Followed by:
Bro, it’s driving me nuts! It does not understand subtleties like ChatGPT does!
Cowork might be amazing but Chat sucks!
First, I laughed.
Claude hype is real, and misleading people in thinking you need to spend 50$ a day to have AI do anything useful for you.
That being said, one word stood out: subtleties.
And, to be fair to Claude, there is a practical reason why that might be the case.
Most of us have been using ChatGPT for several years, before even knowing about Claude.
Back in August 2025, 60% of AI chatbot usage was ChatGPT (US numbers). Claude? 3.5%…
Although cross chats context was enabled quite recently, we have had A LOT of chats with ChatGPT, allowing it to learn how we talk, think, react, etc. One could say it knows our subtleties…
So when people argue one model is better, it might just be because they’ve had more meaningful conversations with this provider, not that the model is genuinely better.
2 takeaways from that:
If you want unbiased chats, you can disable context across chats - labelled as Memory
If you want more aligned chats - that get better over time - make sure you enable Memory
You can also see what each provider knows about you… It is pretty impressive (scary?)
