My intern asked me this on his last day. He is a second-year student. The thing I spent my twenties learning, turning an idea into working software, now takes him a few days.
He wanted to know where he should put the years instead. This is what I told him. Five things, each with something you can do this week.
1. Learn the boring fundamentals. They are what catch the expensive mistake.
The week before he left, I was fixing a bug in Revolv. One person had turned into three records. One had the email, one had the phone number, one had the meeting history. The app treated them as three strangers.
That is a first-week database problem. One person, many attributes, one identity. It looks outdated next to the newest model, until it costs you a week.
AI wrote that code fast. It could not tell me the key was wrong. It would have happily built ten thousand lines on top of it.
This week: if you are in a database or networking class, treat it as the class that lets you see what the tools cannot. If you are not, take one.
2. Build something small enough to fail honestly.
I gave him homework. There is a board called the ESP32, about the size of an AirPod, with a USB-C port and a radio. It costs a few dollars. People are using coding agents to turn them into voice assistants, toys, sensors.
It does not need to become a company. It needs to answer five questions. Does it solve a specific problem? Can someone use it without you explaining it? Would they use it twice? Do you understand how it works? What did you learn when it broke?
A pitch deck can hide a bad idea for a year. A thing on your desk that nobody picks up cannot hide anything.
This week: order the board. Or pick any project you can finish in a weekend and put in front of one person.
3. Use AI. Do not hand it the wheel.
Years ago I got into a taxi in New York. The driver was new to the country and trusted the map completely. It took us toward the Hudson River. I could see the water. He could see the screen. I told him to turn around.
That is your relationship with a coding agent. It is relentlessly encouraging. You can write thirty thousand lines in a week, feel unstoppable, and be spinning. My intern admitted he had felt that. So have I.
The tool is exactly as smart as the question you asked it. Ask a lazy question, get a confident, well-formatted, wrong answer.
This week: before you accept anything an agent builds, say out loud what assumption it made. If you cannot, you are the taxi.
4. Pick the internship where you touch the real thing.
My first internship was an office assistant job. I spent it at a photocopier. I went to learn how a company ran. I learned how a hundred-page document ran through a machine.
A famous name helps a résumé. Participation builds ability. Ask these before you say yes:
- Will I build or change something that ships?
- Will someone explain how decisions get made, not just what they are?
- Can I ask a question without becoming a problem?
- Will I get feedback that costs the giver something to write?
- Will I see how my work lands on another person?
Two or fewer yeses, keep looking.
This week: put those five questions in your notes. Use them in every interview.
5. Check whether you are moving or just busy. Then be kind about the answer.
Do not schedule reflection for Sunday at eight. Do it whenever there is a quiet moment. Am I moving this week, or only busy? What did the last failure teach me? Which question am I avoiding? Do I still care about this problem?
Then be passionate about the goal and gentle about the miss. Those are different muscles. You can be hard on yourself forever and never get there. Passion is rarely the missing piece. Steadiness is.
This matters more with AI in the room, because the tool never says you are spinning. It says great job. You have to be the one calm enough to disagree with it.
This week: one quiet check, honest answer, no punishment.
When my intern's summer ended, I told him the door stayed open. A class raises a question, ask. The board does something strange, show me. Stuck, say so.
Same offer to you. Build something from this list and send it to me.
The internship had an end date. The guidance did not.
