In the early 2000s, I visited my childhood neighbor at college. It turned out that he was smarter than I had given him credit for and was studying computer science at Dartmouth College. The few days I spent there involved learning a considerable amount of specialized ritual and terminology surrounding life in a Dartmouth fraternity. But one piece of folklore has stuck with me in particular since then: the often-heard phrase “Blitz you later.”

BlitzMail was Dartmouth’s homegrown email system. It was just email with a few modifications, really, but nobody called it that. It also had a distinctly non-email tempo and user experience. At a time before smartphones, it was a technique for quick messaging—a reminder, an update to a plan, a flirt. There were computer stations everywhere across campus, often at standing level. Students would log into a terminal while moving between classes, jot out a quick message, and then keep on walking.

The feeling of eerie ubiquity was something like, well, moving through my own campus now and seeing ChatGPT open on so many students’ laptops.

Pick your Blitz

What impressed me at the time was not just the user-space but what might now be called the developer ecosystem. Students were active participants in co-creating BlitzMail. They contributed to the core system, and they built various clients that other students could use to access their Blitzes. Even my neighbor tried building a Mac OS X-native client until other aspects of campus life pulled his attention.

WebBlitz, for instance, was created by a student collective called The Basement, which formed in 1998. In addition to the BlitzMail client, they also maintained a student directory, a marketplace for used stuff, and an election platform for the Student Assembly.

Then there was NetBlitz, which supported easier off-campus access. When it broke down, six years after its original developer graduated, users mourned its loss, and the conservative student paper complained of institutional bias against it. According to reportage at the time:

“I used NetBlitz exclusively, because the Internet firewalls in our housing here in London don’t allow us to get on traditional Blitz,” said Owen Roberts ‘09 in a e-mail.

Roberts is currently studying in London on the Government FSP. He said that he was very upset to have to switch to WebBlitz, as he found it much more difficult to use.

“NetBlitz was faster, more logical, more aesthetically pleasing, had more reliable sign-ins, and didn’t have all of the bugs that WebBlitz seems to have,” he said.

His roommate David Schmidt ‘09, also in London on the History FSP, said that WebBlitz lacks some key features present in NetBlitz.

“[In WebBlitz], it’s harder to tell whether a blitz is sent to you personally or not. On NetBlitz they would bold the subject of blitzes sent just to you. I miss that,” Schmidt said in an e-mail.

BlitzMail was a local implementation of a global email network, and the local version couldn’t keep pace with the pressures of the far larger network. The campus shut down its own servers in 2012 and transitioned to a still Blitz-branded Microsoft suite. But that was only after shaping how a generation of students first experienced online life. (An additional afterlife: I am an avid user of a totally unrelated BlitzMail, a little Android app for sending yourself notes by email.)

I share all this with the caveat that computing at Dartmouth was not all pong games and rainbows. Or maybe it was too much pong. As Joy Lisi Rankin has written, Dartmouth’s fraternity system allied with the leadership of its Computation Center to help build the cultural artifice known as the tech bro. Women involved in early computing were sidelined, and a programming became a macho-coded activity.

Campus cultures matter. It’s no accident that the biggest tech companies in Silicon Valley house themselves not in standard-issue office buildings but boutique corporate campuses. The college campus is a breeding ground for norms that spread elsewhere. Just as Greek life became a template for male-dominated “meritocracy” in tech culture, experiences like BlitzMail’s open ecosystem gave a few generations of engineers a taste of open innovation, resulting in such wonders as the open-source movement. What we do on our campuses now is surely also shaping our relationships with technology to come.

Wasn’t this supposed to be about AI?

It’s very 2026 of me to use a headline about AI to trick you into reading a post on computer history. But here’s where the payoff happens.

My campus, like many other campuses lately, has offered as proof that it is embracing the supposed revolution of generative AI by inking a big, embarrassing, exploitative contract with a major AI provider—which my colleagues and I have at least succeeded in delaying. We bought ourselves some time with our resistance. Time is precious for doing any real thinking. What shall we do and demand instead?

Some of my colleagues would prefer that the campus not support or encourage AI tools in any way. I understand that urge, and there are certainly moments when I feel it. But I also think this technology has some meaningful usefulness, and is weaving its way into society to such an extent that ignoring it means ceding power over it to others.

Assume with me for a moment that generative AI is a technology of at least as much importance as email, and that a college campus should have some provision for it. Join me, also, in the further assumption that how we provision AI is not just a technical matter but a pedagogical one, and a society-shaping one at that. What is to be done?

One strategy is for universities to create models of their own, just like Dartmouth created its own variant of the email protocols. This is what has happened in Switzerland, where the government and research universities joined forces to create Apertus, a foundational model derived from fully documented designs and data sources. (I’m on the board of Metagov, the fiscal sponsor of the Public AI initiative that has partnered with Apertus, and I consulted some with the founders of the consumer cooperative you can join to use the model.) With efforts like this, universities can demonstrate that stealing data and surveilling users—as the big AI firms have done—is not the only way to build a model. Higher roads are possible, and public institutions should be competing with each other (or collaborating!) to build their roads higher. Students can experience AI models as not just something someone else does for you but something that can happen in your own community.

A less expensive starting point would be to host open-access models on university infrastructure—perhaps running on renewable energy and tweaked for efficiency. Like Moxie Marlinspike’s Confer.to and Proton’s Lumo, the inference could be encrypted so as not to retain any student data. Students could help maintain the servers and decide how to balance, for instance, environmental impact with speed and capability. They could also build custom interfaces with the models, share them with each other, and encourage adoption for pocket-change or glory.

There is already a growing middleware economy for AI tools. Software developers use platforms like Open Router and Ollama to manage relationships with multiple models. Less-technical users can try tools like Duck.ai and the aforementioned encrypted platforms. For simple tasks, I often just use bots hosted locally on my laptop or phone. These each have elements of what Ethan Zuckerman has elegantly called “loyal clients.”

We are in the early days of a consequential new technology, and how we introduce students to it can make a real difference in that technology’s future. I hope that we can use this as an opportunity to teach some lessons:

• Technology can be a web of relationships, not just a monopoly product. When students have the chance to learn how their technology works, or even just know the people who do, it changes their experience of technology. I have called this “slow computing.”
• Own the tradeoffs. Every design comes with tradeoffs. Rather than letting a far-off company decide how to balance environment, privacy, user experience, and data ethics, students should have a chance to experience the tensions for themselves.
• Local culture can guide global technology. Just as Dartmouth developed its own terminology and norms around email, our campuses can enable students to experiment with distinctive ways of relating to AI.

Collective governance of AI is happening in many parts of the stack, and it can happen on campuses. I would be proud for a visitor to our university not only to have to endure our peculiarities of sports fandom and revelry, but to experience an entirely different version of AI than what they are seeing elsewhere.