This is a linkpost for my Harvard Crimson op-ed for its commencement issue.
I will not reproduce the whole text here, but my advice to the class of 2026 is in the following parts:
My advice for the Class of 2026 is to embrace AI as a technology, but treat it critically as citizens.
…
Throughout your time at Harvard, you received mixed signals on whether AI was to be avoided or carefully used in approved sandboxes. My signal to you is clear: make sure to master AI and use it as much as you can. Whether your degree is in English or Computer Science, you should be using not just ChatGPT, but also Claude Code, Codex, and products from cool startups I haven’t heard of. Ask yourself how you can use AI to realize ambitious projects that would have been impossible last year, or even last month.On the other hand, using AI does not mean exempting the companies that build and use it, as well as the governments responsible for regulating it, from account. You have a right and a duty to make your voice heard in how this technology should be used to maximize benefit for humanity. You should educate yourself on both its benefits and risks. In my view, some of the latter (such as water usage) are overblown, while others (such as concentration of power) receive too little attention. But you should make your own mind, and even if you oppose AI deployment, still use AI to advance your case.
AI can be extremely empowering for those who possess initiative, creativity, and vision, even if they lack experience or capital. I believe that over the coming years, we will see teams of a few young people achieve results that previously would have required large companies with hundreds or thousands of employees. I wish for you to be one of those people.
Windows On Theory 
I assume you know it, but just to be clear: regardless of the rest of your post (which thankfully is more nuanced), you must realize that the title does not look good on you (and OpenAI)? And that its title is what sets the first and quite durable view of a reader on a piece?To give you an idea, this is how it reads:
“The stuff I’m happily trying to build will kill most of you! Try to survive.”
I know the title is provocative, but it’s a metaphor, and clearly the article is about economic adaptation rather than physical survival. I decided to write this after talking to Harvard students and realizing many of them are not using coding agents, and basically just the chatgpt version that they get from the university. I do believe that it is important for students and young people to adapt to AI and make sure they utilize it.
Dear Boaz,
I’ll start by noting that I’ve been a fan of your work since being inspired by the copy of Arora+Barak that I picked up as a first-year graduate student. We all read many positive and negative takes on AI every day, but I feel inclined to respond to yours precisely because I respect you and because you’re a senior member of our community.
I believe that your post is tone-deaf and perhaps even dangerous. First off, everyone—certainly including students—is very well aware of the impact AI is having. To have a professor write a seeming public service announcement saying, “You need to use it more. MORE! NOW! DON’T GET LEFT BEHIND!” isn’t doing much except projecting a certain kind of mania.
For most people economic adaptation (that is, being able to earn a living) *is* physical survival. Many of my students, including very good ones, have been having quite a hard time finding jobs recently. At the same time, according to Fortune the average OpenAI employee gets $1.5 million in total compensation (source: https://fortune.com/2026/02/18/openai-chatgpt-creator-record-million-dollar-equity-compensation-ai-tech-talent-war-career-retention-sam-altman-millionaire-staff/). I find “learn to vibe code, bro” just as condescending a response to the hardships of computer science students now as I found “learn to code, bro” 10 years ago as a response to the hardships of tradespeople.
Or maybe “just” vibe coding isn’t enough? I find this advice very sad. I believe that higher education should be about encouraging students to learn, create, and explore, and not about encouraging them to farm these things out to AI. I believe that you will do your students the greatest service by sharing your vast knowledge of cryptography, complexity theory, and the foundations of machine learning. No one needs to go to Harvard to learn about vibe coding, coding agents, or token maxxing. It’s possible that students will have a hard time finding jobs regardless, but I think having actual computer science expertise will give them a much better shot.
Finally, I agree with your comment about the concentration of power. With no one able to code or do anything else except prompt AI, we’ll all just believe whatever ChatGPT, Claude, and Gemini tell us. Or whatever their creators get them to tell us. What could go wrong?
Sincerely,
Huck
P.S. This is a 100% artisanally human-written comment.
Dear Huck
I decided to write this op Ed after a conversation with a Harvard senior who told me none of his non-CS friends are using more advanced AI than basic chat.
if you read beyond the title you would see that I am actually optimistic about the ability of young people to both make a living and have an impact.
As I say there, I believe that times of transition favor the young. But to take advantage of that young people will need to think how to use AI to amplify their impact, and they can not count on the University (which has not itself adapted) to teach them that.