I Knew It Was a Virus. I Opened It Anyway.
Christian Muñoz ·
When I was a teenager, the internet felt like another dimension. It was not simply a collection of websites or computers connected to one another. It felt like a place, a vast universe that existed somewhere beyond the screen, waiting to be explored. Perhaps that is why I still have such fond memories of Netscape Navigator, with its stars and ship’s wheel. Even the name suggested what using the internet felt like at the time. You were not browsing. You were navigating. You were going somewhere.
It felt like traveling.
Somewhere during those travels, I found something called “Virus Collection (Mac).” I do not remember exactly where I found it. It may have been AOL, Usenet, a forum, or one of the many strange corners of the early internet. I do remember that it was a large collection, and I remember some of the names inside it: nVIR, INIT 29, and others that have become hazier with time. One, however, has stayed with me for more than twenty-five years.
Garfield.
It even had a little black-and-white Garfield icon.
I knew what computer viruses were, and I understood that opening one was not something I was supposed to do. I also knew what was on my computer. There was nothing irreplaceable on it, and I had backups. Without realizing it, I already had some basic principles of computer security in my head: know what you are risking, keep copies of what matters, and have some idea of how you are going to recover when things go wrong.
Compared with names like nVIR, Garfield seemed almost friendly.
So I opened it.
Deliberately.
I was not tricked. I did not think I was downloading a game or some useful application. I knew it was a virus. That was precisely why I wanted to open it.
I had to see what happened.
More than that, I wanted to understand how it worked. How did something go from being a piece of software someone had built to something that had infected a computer? What happened between those two states? How did it enter the system? Why did it behave the way it did? Was there a pattern? If something could exploit the computer in one way, why could it not be stopped another way?
Not knowing made my brain go into overdrive.
What could happen?
I also knew enough not to expect the virus to announce itself. It probably was not going to display a message saying, “Hey dumbass, you opened a virus!” There would be no welcome screen asking me to accept the Terms of Service and press OK to confirm that I wished to infect my Macintosh.
I double-clicked it.
It opened.
Nothing happened.
That was it.
For a moment, it was almost disappointing. There was no skull and crossbones, no dramatic animation, no Hollywood version of a computer infection. The Mac was still sitting there, apparently unchanged, so I went back to using it.
Then strange things began to happen.
The menus stopped behaving normally. Things that had always worked suddenly did not. I also remember the display eventually looking strange, as though everything had become larger and blurrier, although after so many years I cannot say with certainty whether that was caused by Garfield or something else I had done to that poor computer. I remember the classic Macintosh bomb screen appearing too.
The interesting part was not that the computer had suddenly become possessed. It was that the changes emerged gradually through ordinary use. Something invisible had changed underneath an interface that still looked familiar.
I remember thinking, in a way, “That’s it?”
But that was also the point.
I had wanted to know what a computer virus actually did, stripped of all the dramatic imagery surrounding the word “virus.” Now I was watching the answer unfold in front of me. A system I thought I understood was behaving differently because something inside it had changed.
Eventually, curiosity collided with inconvenience.
Fortunately, I had a great computer teacher in high school. I learned about networking, remote access, file management, disk space, bytes, and even how to crimp network cables. I knew enough to start trying to recover the machine. I zapped the PRAM. I reformatted the drive. I restored everything from backup. If my memory is correct, I ended up reformatting it more than once before everything was back to normal.
I did not deliberately infect the computer again.
It was not because I had lost my curiosity.
Restoring the damn thing simply took too long.
Had I owned a spare Macintosh whose only purpose was experimentation, I probably would have continued.
More than two decades later, I still recognize that teenager in myself. The questions have changed, and the systems have become vastly more capable, but the instinct has not disappeared. Today I find myself asking questions about artificial intelligence that come from much the same place. What is the difference between a token and a session? Where does identity management end and authorization begin? What does containment actually mean for a model? What can an AI agent access? Who decides what it is allowed to do? Most importantly, where are the boundaries, and what actually happens when those boundaries fail?
The difference is scale.
Opening a virus on my Performa 5200CD might ruin my afternoon and force me to restore a hard drive. The capabilities, tools, connections, and speed available to modern AI systems exist on an entirely different level. Curiosity still matters, but so does understanding the consequences of satisfying it.
I am less interested now in breaking things simply to watch them break. I want to understand where they can break. Where does a system stop behaving according to our assumptions? Where are the points at which one seemingly reasonable decision combines with another and produces something nobody intended? How can we find those boundaries before the consequences matter?
If I had a spare computer today, I would happily experiment there.
If I had a spare planet, even better.
Unfortunately, we only have the one.
That same curiosity eventually became part of the inspiration behind Keel, the company I am building today. As AI systems become capable of taking actions rather than simply producing information, I keep returning to a familiar question: what happens after we open the thing and let it run?
I do not want to make experimentation impossible. Quite the opposite. I want people to be able to experiment, build, explore, and discover what these systems can do without every mistake becoming irreversible. Boundaries are not interesting because they prevent exploration. Good boundaries make more exploration possible.
If things become too fragile to touch, there is no fun in trying anything at all.
Perhaps that is the thread connecting a teenager sitting in front of a Performa 5200CD to what I am doing today. I still want to know what is on the other side of the double-click. I still want to understand why something behaves the way it does rather than simply accepting that it does. I still believe that understanding how things fail is part of understanding how they work.
The internet no longer feels quite like another dimension. The stars and ship’s wheel of Netscape Navigator are long gone. The unknown universe I once explored through a dial-up connection has become part of ordinary life.
But every once in a while, something new appears and gives me that feeling again.
There is something out there I do not completely understand yet.
I want to see how it works.
I just make better backups now.