A few weeks ago, a friend of mine who works in traditional finance challenged one of my assumptions about AI.
We were discussing the latest model releases.
New benchmarks.
New reasoning capabilities.
The usual conversation.
Then he asked me a simple question:
"If a better AI model comes out tomorrow, would you switch immediately?"
I answered without hesitation.
"Of course."
But later that evening, I decided to test my own answer.
I opened another AI assistant and tried recreating a workflow I'd been running for months.
Within minutes, I became frustrated.
Not because the model was bad.
In fact, some responses were arguably better.
The problem was something else.
It didn't know the projects I'd been working on.
It didn't know my writing style.
It didn't know the decisions I'd already made.
I spent more time rebuilding context than actually working.
That's when I realized something.
For years, technology companies built moats through network effects.
Facebook became valuable because everyone was already there.
Visa became valuable because everyone accepted it.
But AI may be introducing a different kind of moat.
I don't use an AI because other people use it.
I use it because it already understands me.
The hidden weakness in today's AI race is that most people are still focused on intelligence.
Yet intelligence is becoming cheaper every year.
Context isn't.
This is why OpenGradient feels interesting.
The project appears to be built around an assumption that many people still underestimate:
The future winner may not be the AI with the highest benchmark score.
It may be the ecosystem that owns the richest context layer.
The controversial implication?
The next AI monopoly may not be built on models.
It may be built on memory.
My thesis:
Context creates stronger lock-in than network effects.#opg $OPG $BSB @OpenGradient
