She Coded the Boss, Then Demolished It: The Teen Who Rewrote the Rules of Speedrunning

Speedrunning has always had a complicated relationship with the word "legitimate." Glitches get debated. Tool-assisted runs get their own separate leaderboards. Emulator accuracy gets argued about in threads that stretch to 400 posts. The community is genuinely, passionately obsessed with drawing lines.

Naomi Chen just drew a line nobody knew existed — and then sprinted across it.

The 17-year-old from the Phoenix metro area has been running Null Protocol, a mid-budget indie robot combat game, since she was 15. She holds three category records. She's well-liked in the community, known for detailed route documentation and a dry sense of humor on her streams. She is, by any measure, one of the better Null Protocol runners alive.

She is also, depending on who you ask, either a genius or a cheater — because in March 2025, she reverse-engineered the game's primary AI boss, rebuilt it from scratch with optimized behavioral patterns, modded it back into the game, and then ran a flawless sub-22-minute any% that the existing leaderboard has absolutely no idea what to do with.

What She Actually Built

The boss in question is VANCE-9, a bipedal combat machine that serves as Null Protocol's final encounter. In the base game, VANCE-9's behavior tree is notoriously chaotic — it uses a pseudo-random aggression system that makes some fight cycles faster and some brutally slow depending on RNG. Top runners spend significant time on "VANCE manipulation," a set of techniques designed to push the AI toward faster attack patterns.

Naomi's approach was different. She pulled the game's decompiled code, mapped VANCE-9's full decision tree, identified every node where randomness introduced time loss, and rebuilt the AI with deterministic behavior — a version of VANCE-9 that follows a fixed, optimized attack sequence every single run.

"I wasn't trying to make it easier," she told us over a call from her family's home outside Scottsdale. "I was trying to make it consistent. There's a difference. The random version is actually harder sometimes. Mine is just predictable."

The resulting run — which she streamed live, with full transparency about the mod — clocked 21:47, nearly four minutes faster than the previous world record. She uploaded the mod files alongside the video. The community response was immediate and extremely loud.

The Leaderboard Problem

Speedrun.com's Null Protocol moderation team has been in a documented holding pattern since the run dropped. The current stance, per a pinned forum post, is that the submission is "under review pending category clarification." That post is now six weeks old.

The community is not waiting patiently.

Arguments in favor of rejecting the run center on a straightforward principle: you can't run a game you modified. The AI is part of the game. Changing it is no different from editing the level geometry or removing obstacles. Former world record holder Derek "Dek_Runs" Palacio has been among the most vocal critics.

"If I built a version of the game where VANCE-9 just stands still, I could finish in 18 minutes," Palacio wrote in a thread that hit 600 replies. "That's not a speedrun. That's a tech demo."

Naomi's defenders fire back with equal force. Modded categories already exist across dozens of games. ROM hacks have their own leaderboards. The question, they argue, isn't whether the game was modified — it's whether the modification represents skill.

"She built the thing she beat," says runner and community organizer Tasha Merritt. "That's not removing a wall. That's a completely different kind of mastery."

A Ranking of Where the Community Actually Stands

Based on forum activity, streamer commentary, and direct conversations with Null Protocol runners, here's roughly where the community has landed:

Firmly Against (approximately 35%): The run should be rejected from all existing categories. A new "modified AI" category could be created, but the record should not stand alongside unmodified runs.

Firmly For (approximately 28%): The run demonstrates a higher level of technical skill than any existing record. It should be accepted with appropriate category notation.

Complicated Middle Ground (approximately 37%): The run is impressive and the mod is legitimate as a technical achievement, but the leaderboard structure genuinely can't accommodate it without creating precedent that opens up messy questions for other games.

Nobody is calling her a fraud. Even Palacio, her most prominent critic, has publicly acknowledged that what she built is technically sophisticated. The argument is structural, not personal — which is, honestly, a more interesting argument.

The Bigger Question Naomi Is Forcing

Here's what makes this story matter beyond Null Protocol specifically: she's not the only one who could do this.

As more games ship with accessible modding tools and decompiled code circulates more freely, the barrier to rebuilding game AI from scratch is dropping. What Naomi did required serious programming knowledge and a deep understanding of the game's architecture. But it didn't require a computer science degree or a professional development background. She's 17 and taught herself most of it through YouTube tutorials and GitHub documentation.

If the speedrunning community doesn't develop a coherent framework for AI modification categories now, they'll be building that framework under pressure when a dozen runners from different games show up with similar submissions simultaneously.

"I kind of expected this reaction," Naomi says, with the particular calm of someone who has thought about a problem longer than everyone currently arguing about it. "I just figured someone had to go first."

She's already working on a second mod. She won't say which game.