Personal Intelligence
What If Your AI Knew You for 10 Years?
Exploring the difference between conversation history, and context.
Let’s set the scene
Imagine it’s 2036.
You’re preparing to make an important career decision. You’ve been offered a leadership position at a new company, but accepting it would mean leaving a team you’ve spent years building.
You open your AI assistant and ask:
“Help me think through whether I should take this opportunity.”
Today’s AI could give you a thoughtful answer. It might outline the advantages and disadvantages, suggest questions to ask, and help you compare compensation, responsibilities, and career prospects.
But imagine something different.
Imagine your AI had access to ten years of your professional development: projects you completed, decisions you made, mistakes you learned from, goals you pursued, and even the goals you eventually abandoned.
It could retrieve a journal entry from 2029, when you accepted a promotion for reasons that seemed compelling at the time. It could connect that decision to a reflection you wrote two years later about what you wished you’d considered.
It might notice that your priorities have shifted. The things you valued at thirty aren’t necessarily the things you value at forty.
Instead of simply offering general career advice, it could help you examine the decision against the documented history of your own experiences.
Not because the AI had become more intelligent than every other AI.
But because it had access to something no other AI could manufacture: the context of your life.
This raises an intriguing question.
What would ten years of accumulated personal context actually make possible?
And perhaps more importantly, would remembering ten years of information really make an AI better at helping you?
A decade of conversations isn’t a decade of understanding
Modern AI systems are extraordinarily capable. They can explain complex subjects, write software, analyze documents, and help us work through difficult problems.
Yet much of our interaction with them remains strangely repetitive.
We explain our backgrounds. We describe our projects. We establish our preferences. We provide the same context again and again.
Some systems now preserve information between conversations, but remembering details and understanding their significance are different problems.
Consider three concepts: memory, knowledge, and understanding.
Memory allows information to be stored and retrieved.
Knowledge organizes information into facts, concepts, and relationships.
Understanding allows that knowledge to be applied, questioned, and transferred into unfamiliar situations.
Human cognitive research illustrates this distinction.
In a well-known 1981 study, researchers Michelene Chi, Paul Feltovich, and Robert Glaser examined how physics experts and novices categorized problems. Novices tended to group problems by their visible characteristics. Experts were more likely to recognize the underlying principles connecting problems that looked different on the surface.
The difference wasn’t simply that experts remembered more equations. Their knowledge was organized differently.
Now consider an AI containing ten years of your conversations.
It might retrieve the exact day you decided to learn programming. It might find a discussion about a difficult project or recall a preference you expressed years earlier.
That’s useful. But it’s primarily retrieval.
A more ambitious system would connect those records.
It could recognize that a challenge you’re facing today resembles one you encountered years ago, even when the subjects appear unrelated.
It could distinguish between something you once believed and something you later discovered was wrong.
It could identify not only what you’ve learned, but how your understanding developed.
None of this requires imagining an AI with human consciousness or a human-like understanding of your identity. It requires better ways to organize, connect, update, and reason over personal information.
And that is a considerably harder engineering problem than saving conversations.
What accumulated intelligence could actually look like
To understand the possibilities, consider three ordinary parts of someone’s life: learning, professional development, and personal projects.
Learning that builds on your actual experience
Suppose you begin studying statistics in 2027.
During your first year, you struggle repeatedly with a particular concept. You work through examples, make mistakes, and eventually find an explanation that makes the idea click.
Years later, you’re studying machine learning.
A conventional AI tutor can explain the new material. A system with your learning history could also retrieve your earlier difficulties, recognize relevant prerequisites, and suggest revisiting concepts you previously found challenging.
By 2036, perhaps you’re teaching the subject yourself.
Now your historical context becomes valuable in an unexpected way.
You can ask:
“What did I find confusing when I first started learning this?”
The system can retrieve your actual beginner questions, mistakes, and breakthroughs.
You’ve gained expertise, but you’ve also lost some of the perspective you had as a beginner. Your earlier records preserve a view of the subject that may now be difficult to reconstruct.
Ten years of context wouldn’t merely help you remember what you studied.
It could help you understand how you became knowledgeable.
That doesn’t mean personalized AI tutoring is automatically superior. Its benefits would need to be measured against good teaching methods available to everyone. But a documented history of your own misconceptions offers information a new assistant simply doesn’t possess.
Professional experience that doesn’t disappear
Consider how much knowledge disappears into meeting notes, abandoned plans, project retrospectives, and forgotten decisions.
You might spend years making estimates, choosing technologies, managing projects, and evaluating opportunities.
But how often do you systematically compare your predictions with what actually happened?
Imagine asking:
“Over the last five years, which kinds of projects have I consistently underestimated?”
If you’ve recorded both estimates and outcomes, an AI could analyze your actual history rather than offer generic advice about project planning.
Or consider:
“What changed my mind about remote work?”
A system preserving dated decisions and reflections could reconstruct how your position evolved.
It might reveal that an experience in 2028 shaped a decision in 2031, which influenced a policy you adopted in 2034.
The value isn’t merely remembering what happened.
It’s preserving the reasoning behind decisions before hindsight changes the story.
Of course, the quality of that analysis depends on the quality of the records. A system containing hundreds of plans but almost no outcomes cannot reliably tell you which decisions worked.
Accumulated intentions are not the same as accumulated experience.
Personal projects that retain their history
Now consider something less abstract: renovating a house.
Over ten years, you replace a roof, remodel a kitchen, install new heating equipment, repair plumbing, and repaint several rooms.
Along the way, you accumulate receipts, warranties, photographs, contractor information, measurements, and notes.
In 2036, you ask:
“Which warranties are expiring next year?”
Or:
“What was behind the kitchen wall before we closed it up?”
Or:
“Did replacing the heating system actually reduce our energy costs?”
A capable AI could answer these questions if it had the relevant records.
But a fresh AI cannot invent the photographs you took in 2028, the equipment you installed in 2029, or the utility bills you’ve collected since.
This is an important distinction.
Future AI models will almost certainly become more capable at reasoning and analysis. But greater intelligence alone cannot recover personal information that was never preserved.
The most interesting opportunities may emerge when more capable AI models meet a durable history of personal knowledge.
Who should own your accumulated knowledge?
Now imagine you’ve spent ten years building this personal context.
Your assistant has access to your learning history, professional decisions, projects, preferences, and the connections between them.
Then the company providing the service changes its pricing.
Or discontinues the product.
Or introduces a feature that requires you to surrender more control over your information.
What happens to those ten years?
This is one of the most important questions in the future of personal AI.
A decade of accumulated knowledge isn’t merely an account history. It may become a valuable intellectual resource: a record of how you learned, what you created, and why you made particular decisions.
If that resource exists only inside a proprietary platform, leaving the platform could mean losing much of its value.
The more useful the accumulated context becomes, the more difficult switching providers may become.
That’s an unusual tension.
A system designed to empower you could simultaneously make you increasingly dependent on the company operating it.
There’s an alternative worth considering.
What if the underlying knowledge remained in files and formats you controlled?
What if your notes, documents, and records survived changes in AI providers?
What if the intelligence layer could evolve independently of the information it worked with?
Under that model, the AI isn’t the owner of your accumulated knowledge.
It’s a tool for accessing and using it.
For something intended to last ten years or longer, durability matters.
The AI model you use in 2036 will probably be very different from the one you use today.
Your knowledge shouldn’t have to disappear every time the technology changes.
This is why Yodis was designed to create the seed that gets planted on your local machine, allowing the tree of your personal knowledge to grow locally.
You control your data. If you want to eventually take that tree and run it with another AI service, the contents you’ve built with Yodis allows that seamless transition.
The difference ten years could make
Return to that career decision in 2036.
You have an opportunity in front of you and a decade of experiences behind you.
A conventional AI assistant can help you evaluate the opportunity using general knowledge and whatever information you provide in the moment.
An assistant with ten years of well-maintained context might do something more.
It could retrieve the reasons behind earlier decisions.
It could show how your priorities changed.
It could identify recurring patterns in your choices, while acknowledging that patterns aren’t necessarily proof of cause and effect.
It could remind you of questions your earlier self wished you’d asked.
And it could challenge your assumptions using evidence from your own history.
Instead of asking how much information an AI can retain, we should ask how effectively it can preserve, organize, question, and apply the information that matters.
And perhaps we should measure its value not by how familiar its answers feel, but by whether it helps us learn more effectively, make better decisions, and avoid repeating mistakes.
This is one of the questions we’re exploring as we build Yodis.
We believe personal knowledge should remain under the control of the person who creates it, and that AI has the potential to become more useful when it can work with that knowledge over time.
But the promise of accumulated context is something to investigate, not something to take for granted.
Ten years from now, the most valuable AI might not be the one that remembers every conversation you’ve ever had.
It might be the one that can help you recognize how far you’ve come, understand why your thinking changed, and see possibilities you would otherwise have missed.
Not simply an AI that knows more about the world.
An AI that can help you make better use of what you’ve learned about it.
Yodis Insights explores the intersection of artificial intelligence, personal knowledge, and lifelong learning. Learn more at yodis.ai.