Zero Experience, Infinite Leverage
When people heard we were 10x'ing our internship program at the same time we were going all-in on AI, I got a lot of questions about it.
Industry thinking goes something like this: AI does the boilerplate, juniors used to do the boilerplate, so interns and new grads are the first to cut.
We are taking the opposite bet. We are going to hire more interns and extend more permanent offers to interns than ever before because we have gone all in on AI.
I want to explain why and tell you what I think has actually changed about starting an engineering career, and what hasn't changed at all.
AI at Shopify
It’s not a secret that Shopify takes AI coding seriously. Tobi’s memo about reflexive AI usage had a massive impact on how AI is used and perceived inside Shopify. If you are starting your own career in tech, you should probably be weighing how well the company is preparing itself, and can prepare you, for the changes we are seeing in our industry.
We’ve built something we call River. River is a coding agent that is different from Claude Code, Codex, Pi or whichever harness you have been using locally. You can think of River as your coding partner who has already been onboarded, knows our codebase well and understands it deeply. River is a senior developer who has all the time in the world for you.
To use River you tag @river in a Slack channel like you'd tag a coworker. It can read code, write code, run tests, query our data warehouse, look at production traces, and open PRs. By design, it does not respond to direct messages. If you DM River, it politely tells you to make a public channel. Tobi wrote about River here and why we made a lot of these decisions.
That public-only constraint is the thing that has had the biggest impact for us and is making the biggest difference in how we approach internships.
In the last month, ~5,938 people across Shopify worked with River across 4,450 public Slack channels. The merge rate on River's PRs to production went from 36% to 77% over two months, and almost none of that improvement came from the underlying model. It came from people watching River work, noticing where it got stuck, and writing down what it should have known. The agent got better because the company got better at using it, in public, where everyone could see. What started as a single-player tool evolved into a multi-player activity.
That is the part I want new engineers to understand. River is not just an agent, it’s the thing that will help accelerate your internship and career faster than has been possible in the past. We are investing heavily in it and I think that is a big opportunity for you.
Mentorship we can all use
I have run internship programs at small companies and large ones for several decades now.
The thing interns need most, and has always been the hardest to get, is exposure to real work, real feedback, a real person who can explain why something is the way it is. Because of that, the bottleneck for internships has been the time of senior engineers.
Before River, if you were a new engineer trying to learn how a senior person scopes a problem, you had to either be in the room or wait for them to write something down. Now you can scroll through #tobi_river and watch our CEO talk through a problem with an agent in real time. You can see me working in my own #farhan_river every day. You can read how a backend engineer in a totally different part of the company convinced River to construct the right log query, then try the same prompt the next day on your own work.
When I was an intern, learning by watching meant pair programming next to one person for an afternoon. That kind of interaction and mentorship is still the gold standard, but now it is supported by being able to read thousands of public threads where those same experienced engineers are thinking out loud, in front of you. You’ll see them make genius moves, stupid mistakes, and even ask for help.
You still have to go and get it
In my experience, mentorship at Shopify is most useful to people who actively go and get it. River makes it more visible, but just seeing does not become learning on its own. You still have to read the thread. You still have to ask the follow-up question in the channel. You still have to try the technique on your own work and notice when it fails.
In the cohorts I have worked with, the interns who grow fastest tend not to be the ones with the most conventional resumes. Successful interns are the ones who treat being new as an asset. They ask questions early, take the harder problems when given the choice, and then they put in the reps.
The reason this now matters more, not less, is that when the cost of producing code drops, the value of being the engineer who actually understands what got produced, and why, goes up. I think curious engineers have way more leverage than they used to.
I have written before about the "hard path" and how I think it tends to pay off even when it doesn't work out, because you come out of it with skills and relationships. I would say the hard path is now easier to start but just as hard to finish. AI gets you to the first prototype quickly, it gives you options to choose from, and it even does entire pieces of work all on its own, but it doesn’t replace the curiosity, creativity, and judgement that comes with human experience.
Why now might be the best time to be starting your career
Three things make this moment in time special.
First: you already reach for new tools. Most people in the industry have habits to unlearn. They have to "adopt AI," which is a polite way of saying they have to fight their own muscle memory. You don't. AI isn't a new tool bolted onto an old way of working for you. It's just how work gets done. You're AI native. We saw the exact same thing a decade ago when we hired hard out of the mobile-first cohorts to build mobile apps. We didn't want people who were adapting to mobile. We wanted people for whom mobile was the only thing they'd ever known. That's you, now, with AI.
Second: you ask why. "Why are we doing it this way?" And you ask it with genuine curiosity, not as a critique. You'll take a chance to step up and build the prototype everyone assumed was too ambitious to try. Last year one of your peers deleted six lines of code and saved Shopify roughly $600,000 a year in infrastructure costs. Six lines. No activity metric on earth would have flagged that as a productive day. Whatever the situation, a “beginner’s mind” is now hugely valuable.
Third: if you’ve done an internship, you've already done the interview. A four-month internship is the best technical interview I know. Eight hours of interviews tells me something. Four months of real work, in our codebase, with our review process, on a real team, tells me almost everything. We convert a meaningful number of interns to full-time, and by the time we make the offer we already know, with high confidence, whether the fit is there. When you've been working in public for four months, you don't have to sell me your story. The work already told it.
How we evaluate technical talent now
We are still figuring this out. The framework we have been borrowing from the University of Waterloo has three modes:
- No AI allowed. Can you write the code by hand? Do you understand the layer below where you are working?
- AI optional. Do you know when to reach for it and when to push your own thinking further?
- AI mandatory. The problem is too big for the time available. Can you scope it, prompt it, validate the output, and ship something that works?
I like this framework, and I'm sure it will need to evolve. The biggest open question is whether candidates who score well on these techniques actually perform well on the job over a year or two. I don't think anyone in the industry has cracked that yet. We are watching our intern cohort closely to find out.
One thing we have not changed: whoever submits a PR owns the code in it. You can use any AI tool. A human reviewer still puts their name on it, and so do you. AI does not currently get to merge code at Shopify on its own.
Why Shopify specifically
I'm not going to argue that Shopify is the only good place to start an engineering career. But I will argue that for a certain kind of person, it is absolutely the best place to come right now.
We have pair programming in our DNA. The third in the "pair" is now an agent that is a deep part of our shipping culture and is designed to help you learn. We have a public-by-default communication model. River makes that culture searchable for you. We have a monorepo big enough that you can spend years learning new parts of the codebase, and a deploy cadence aggressive enough that you ship in your first weeks. We have leadership that codes: I still pair on interviews, and Tobi is publicly working with River in a channel that anyone can read.
The risks for you that I care about
A few things keep me up at night when I think about someone starting their career.
The first is cognitive atrophy. You might call it “brain rot.” The brain is a muscle. If you let AI do all your thinking, you will lose the ability to do it yourself, and you will notice it the day the model gives you a confident wrong answer about a system you don't understand. My rule of thumb on my team is that you should be able to explain the layer two or three below the layer you are working at. AI is allowed to help you learn. It is not allowed to replace the learning you'll need to do.
The second is the temptation to evaluate, and be evaluated, for productivity by output volume. Lines of code, PRs shipped, tokens spent. None of these tell you whether the work was good. The intern who saved $600,000 a year wrote a negative number of lines that week. Weekly demos are still the most reliable signal we have for whether real work and progress is getting done.
The third is making sure interns get real mentorship and not just access to a tool. River does not replace a senior engineer pairing with a junior on a hard problem. It expands what that pairing can cover in an hour.
If you are early in your career
The advice I give early-career engineers has not changed in thirty years. AI has just removed the last excuses for not following it.
Build something. Not a course project. Not a resume bullet. A real thing that solves a real problem for you or someone you know. Put it on GitHub. Find an open-source project that has issues and try to fix one. If you want to work at Shopify, build something against our APIs. There is more interesting work in the world than there are people doing it. You do not need a job to start.
Then come find us, or come find somewhere else that will let you do real work alongside people who are better than you and willing to explain why. The right place to start a career is the place where you will grow fastest. For a certain kind of curious, intense, AI-native person, I think Shopify is the best option for you right now and I am working hard to live up to that.
We're hiring 1,000 interns again this year. The bet is the same one I would make on myself if I were 22 today: show up where the work is interesting, the culture is ready, and the tools are sharp.
Then put in the reps.