Tech & Leadership
What I'm Telling My Nephew About AI and Our Future Economies
Back in November, I had a conversation with my nephew. He’s entering high school and he’s actively worried that AI is going to take all the jobs in software engineering before he even gets a chance to start.
I wanted to tell him he’s wrong. But I’d just spent the evening watching Claude Code spin up a team of subagents to build a project I’d barely finished conceptualizing. The whole thing worked. And that’s the problem.
This is where I live now. Building with these tools every day while wondering what kind of world my kids are going to graduate into.
The Junior-to-Senior Pipeline Problem
Junior engineers used to grow into senior engineers by doing the repetitive, derivative work, implementing patterns someone else designed, fixing bugs in code they didn’t write, slowly building the intuition that separates a competent developer from a great one. That’s how the pipeline worked. You did hundreds of reps on the fundamentals until the architecture clicked.
But if AI coding tools, the kind that can plan, execute, and iterate on their own, handle most of that derivative work now, where do juniors get their reps?
The risk is: if AI doesn’t reach senior-level programming quality in the next five years, but it does replace the work that grew juniors into seniors, the overall quality of code globally would start to degrade. We’d be cutting the bottom rungs off the ladder while assuming the top stays sturdy.
Now, maybe AI code ends up being better than what juniors would’ve written, and the floor actually rises. I’d love for that to be true. But even in that scenario, we still need people who understand why the code works, not just that it does. And that understanding has always come from doing the work yourself.
And honestly, I don’t think most people in industry have fully reckoned with that yet.
The Economics of Low-Value Work
Here’s where it gets interesting, and a little unsettling.
When a task that used to cost a few hours of an engineer’s time now costs $20 in Claude credits, the math changes. Work that wasn’t worth doing at engineer-salary economics suddenly becomes profitable. Bug fixes, small features, one-off scripts, migration cleanup. All of it gets cheaper, which means more of it gets done.
On the surface, that sounds great. More work gets done. More value created. Efficiency everywhere.
But follow the thread. If we’re automating tens of thousands of lines of code per week, who’s reviewing it? Who’s making sure it all holds together? I know what the counter-argument is: “AI will get so good we won’t need to review it anymore.” Maybe. Eventually. But if you asked any senior engineer in industry today whether we could blindly trust all code pushed by AI, I’d expect an overwhelming majority to say no. That’s not the reality we’re dealing with in March 2026.
The ground is shifting under everyone’s feet, and not just in tech.
Some Honest Advice
I’ve been thinking about what I’d tell my nephew. Or anyone else trying to figure out where they fit in this landscape. Here’s where I’ve landed:
1. Get your hands dirty now
At a minimum, you need to understand what these tools can do and where their limits are. Pick something you’re genuinely interested in, a personal project, a tool you wish existed, anything, and have Claude Code try to build it. Don’t just read about AI. Use it. That hands-on familiarity is worth more than any think piece (including this one).
2. Coding might become like playing guitar
Here’s an uncomfortable thought: writing code may become something closer to playing guitar. Lots of people will do it. It’ll be a valuable skill. But only an extraordinarily select few will get paid well to do it professionally, not because the market doesn’t need code, but because their performance is that good, there will always be demand for them.
The unfortunate conclusion is more stratification. People who are highly skilled will be rewarded disproportionately, and people who aren’t will have fewer opportunities. It’s the same pattern you see in music, acting, and professional sports. Lots of talented people, but very few actually make a living.
3. Engineering thinking outlasts engineering tools
Breaking down problems. Systemizing messy situations. Organizing chaos into something solvable. This mode of thinking, the actual engineering in software engineering, remains valuable regardless of what industry you’re in or what tools exist.
There’s this exercise from 7th grade science I keep coming back to. The teacher asked us to write instructions for making a peanut butter and jelly sandwich, then followed them literally. Put the jar on the bread, tried to spread with the lid on. Nobody was specific enough. That’s what working with AI feels like. The quality of what you get out is almost entirely determined by how well you decompose the problem before the tool touches it.
The best managers I’ve worked with think this way. So do the best teachers, the best event planners. Really, anyone whose job is to take a mess, articulate what’s wrong, and tackle it step by step. If you’ve trained yourself to decompose complexity, that skill transfers everywhere.
4. For people outside tech: find what survives
If you’re not in the tech industry, the question isn’t “will AI take my job?” It’s “which parts of my job generate value that AI can’t replicate?”
For example, paralegals don’t need to spend weeks gathering and summarizing case documents anymore, ChatGPT does it faster and arguably more accurately. But AI can’t go to court. It can’t build the relationship with a client who needs to trust you with their livelihood.
Or if you’re a digital artist, the market is already flooded with AI-generated work. That actually makes high-quality, hand-made pieces more valuable, not less. But you’ll need to build a brand, find customers who connect with your specific taste, and do the human parts of the business that a model can’t.
Ask yourself: what are all the things that make this thing a business? The answer is usually the human parts.
Closing Thoughts
I don’t have a neat bow to put on this. I’m literally living in two realities at once, dependent on these tools to keep up with expectations that multiply every quarter, while having deeply worried conversations with the people I love about what the world looks like for kids growing up right now.
The Bay Area tech bubble makes it hard to stay objective. For every person that’s excited, there’s another side to that person that’s uneasy. And I think being honest about both is more useful than pretending we’ve got it all figured out.