Reverse engineering
Where I actually learned to build: teaching myself to reverse-engineer games — until Riot's lawyers sent a cease-and-desist.
From about nine to thirteen I taught myself to reverse-engineer game binaries and build cheats — first for Fortnite, then Valorant. It grew into a real operation with actual users, run against professional anti-cheat teams (BattlEye, Easy Anti-Cheat, and eventually Riot’s kernel-level Vanguard) whose entire job was to shut me down. Then Riot’s lawyers sent a cease-and-desist, and I stopped. It’s where I learned to teach myself hard things, ship under pressure, and think in adversarial systems — the instincts I now point at high-stakes AI.
Before I had the words for it
I was around nine when I started pulling games apart to see how they worked. No mentor, no source code, no map — just a disassembler, a wall of forum posts, and enough stubbornness to keep going until it made sense. Long before I knew the phrase reverse engineering, I was doing it: staring at a running process until I understood how the game kept its whole world in memory.
It started with Fortnite and moved to Valorant. And it stopped being a hobby the day other people started relying on what I'd built. It became an operation — something I had to keep alive every single time the developers pushed an update. I was a kid running live distribution to a real user base, against teams of professionals paid to shut me down.
The arms race is where I actually learned to build
I went up against BattlEye, then Easy Anti-Cheat, and eventually Riot's kernel-level Vanguard — systems built by well-funded engineers whose entire job was to end exactly what I was doing. Every week they'd change something; every week I had to understand the change and adapt or the whole thing went dark. Nothing has taught me more about how real systems work, and fail, than being on the losing side of a professional blue-team's attention for a couple of years.
Understanding how defenders think, and how attackers think about defenders, is most of security and a large part of AI.
The letter
Then Riot's lawyers sent a cease-and-desist. Riot doesn't bluff about this — they've taken cheat makers to court for millions, and studios have gone after developers younger than I was. Getting that letter as a teenager was the moment the game stopped being a game.
I stopped. Not because I'd been outsmarted, but because I finally understood I'd been building against people instead of for them. I was a kid; I owned it; I moved on.
What it's worth now
I took the same instincts — take the system apart, find where it breaks, build something that holds under pressure — and pointed them at problems that actually matter. It turns out that intuition for adversarial systems is exactly what high-stakes AI needs: red-teaming, robustness, evals, knowing that anything you ship will be probed and building it to fail loudly instead of silently. The clearest line from that era to now is the half-life of a detector — the thing I lived weekly then, turned into a way of thinking about security and ML now. The full version of the story is here, and where it points is Littman.