In 2021, a young man showed me the DALL-E pilot. His parents are close friends. He'd just taken an internship at OpenAI.
After the demo, I told him straight: based on what I just saw, CS students won't have anywhere to work. Their jobs will be done by this.
He didn't answer right away. Smiled a little and said: "Well, machines still need someone to operate them."
Fair point. Machines do need people. What he didn't convince me of was that we'd need anywhere near as many.
That conversation stuck. Not because he was wrong exactly — but because neither of us had any way to measure what came next.
The thing that annoys me about how people talk about this.
It's not denial. Most people know the storm is coming. The thing they keep getting wrong is the pace. They see the direction, assume they have time, and then get surprised anyway. The velocity is the thing. It keeps catching people off guard — including people who really should know better.
The irony that started this.
Software engineers are building AI. They're also the people closest to the replacement wave. The same hands writing the code are the most exposed to what that code eventually does.
That felt like a signal. If I wanted to understand how people reposition under AI pressure, engineers were the right place to start — not because they're the only ones affected, but because the pressure on them is earliest, most visible, and most measurable.
What I'm actually building.
HumanExodus is not career advice. Not a prediction engine. Not a dashboard.
It's measurement infrastructure.
The atomic unit is:
AI-Induced Pressure → Human Repositioning → Observed Outcome
The goal is an open, longitudinal dataset of how people actually reposition — not what they say they'll do, but what they do, and what happens after.
Here's the honest truth: nobody really knows where AI leads. Not for the economy, not for society, not for culture or ethics. The destination is still forming. But longitudinal data gives us something useful — it lets us see which vectors are emerging to dominate the patterns. That's not prediction. That's at least navigation.
Why open. Why now.
I'm not an engineer. This project exists because I kept watching and couldn't find anyone measuring it properly.
The schema is open. The methodology is open. The data will be open as it accumulates.
If you've repositioned because of AI pressure — changed roles, picked up new tools, shifted focus, left the field — your record belongs here. Not as a data point. As evidence.
HumanExodus: github.com/shenbrian/humanexodus
