The End of One-Size-Fits-All AI

Today’s leading AI models have something in common.

Whether you use ChatGPT, Claude, Gemini or Grok, they’re all designed to answer your questions in roughly the same way.

Sure, each has its own personality and strengths. But at their core, these models are trying to be a general-purpose assistant for everyone.

But I don’t think that’s where AI is heading.

Because the companies building these systems have realized something.

The most useful AI won’t be the one that knows the most.

It’ll be the one that knows you the most.

AI Gets Personal

Last January, I wrote about Lenovo and Motorola’s vision for an “always-on” AI assistant that quietly observes what’s happening around you instead of waiting for a prompt.

Image: Motorola

It was designed to understand meetings, conversations and daily routines so it could help when you actually needed it.

I still believe that’s true. But I also think it’s pointing toward something even bigger.

The next step isn’t just AI that understands your surroundings. It’s AI that understands you.

And that shift is already underway.

OpenAI has expanded ChatGPT’s memory so it can reference information from past conversations and use that context later. In 2025, the company updated memory so ChatGPT could reference chat history beyond a simple list of saved memories.

Anthropic is moving in the same direction with Claude. Its memory tools are designed to store and retrieve information across conversations, building up knowledge over time instead of starting from scratch every session.

Meta is going even broader.

The company says it will begin personalizing content and ad recommendations across its apps based on how people interact with its generative AI features. That means AI conversations could eventually shape what people see across Facebook, Instagram and other Meta platforms.

All these upgrades might sound like small convenience features, but they actually represent the first generation of personalized AI.

And it’s arriving at an astonishing scale.

OpenAI says ChatGPT now serves roughly 1 billion monthly active users.

It’s estimated that Claude draws over 55 million monthly active users.

And Meta AI now also serves more than 1 billion monthly active users.

So it’s not like these are some niche features being tested in a lab. Hundreds of millions of people are already using AI assistants that are slowly becoming more personal.

And improving their memory is only the beginning.

Imagine asking an AI to explain a complicated medical diagnosis.

A generic chatbot might give every user roughly the same answer. But a personalized AI could respond very differently.

If it knows you’re a doctor, it can skip the basics.

If it knows you’ve never studied biology, it can slow down and explain everything in plain English.

And if it knows you tend to get anxious about health information, it can be careful about a diagnosis without being vague.

Same question and same underlying model. But you’ll get a completely different answer based on context.

And that’s what I think a lot of people are missing about where AI is heading.

Personalized AI isn’t just about remembering that you like Italian food or prefer short emails. It’s about adapting the way information is framed, explained and delivered.

And researchers are already working to make that future a reality.

One emerging field is called personalized alignment. The basic idea is that today’s AI systems are often trained to satisfy the “average” user. But future systems will likely need to adapt to the preferences, goals and communication styles of individual users.

That’s a much bigger change than it might sound.

For most of computing history, software has worked the same way for everyone. My copy of Microsoft Word behaved like yours. So did my web browser and spreadsheet software.

But that doesn’t have to be the case with AI.

And that’s why I believe it will become the first major computing platform where every copy gradually becomes unique as it learns from the person using it.

Over time, your AI will learn how you write, how you make decisions and how much detail you want. It’ll understand how you prefer to learn, what risks you’re comfortable taking and what ideas you want challenged.

By learning what values matter most to you, AI will become dramatically more useful. It will become a better teacher, a better health assistant and even a better work partner.

But this level of personalization will inevitably create new problems.

Because the more an AI adapts to you, the more it has to decide whether to challenge you or agree with you.

Should it push back when you’re wrong, expose you to ideas you disagree with or soften hard truths because it knows how you’ll react?

Or should it become exactly the assistant you want, even if it means reinforcing your blind spots?