Before he helped launch Pedalboard Games and later joined indie studio Strange Scaffold, Colin McInerney spent his college years inside one of the most famous game development pipelines in the industry: Bethesda Softworks QA. And his time testing Fallout 4 on Xbox One didn’t just involve ticking boxes and following test plans. One day, he decided to see how far he could push the game until it snapped. The result was a spectacular “super-nuke” rampage that reportedly triggered four crashes in a single morning.
McInerney recently revisited that era in an interview during GDC, explaining how his approach to quality assurance evolved from routine publisher-style verification into a more creative, developer-adjacent kind of stress testing. Instead of only confirming whether features worked as intended, he started hunting for the kinds of failures that happen when players do the unexpected, stack systems on top of each other, and generally treat an open-world RPG like a playground.
That philosophy is also why he pushed back on the idea that generative AI can replace human QA testers. AI can repeat, simulate patterns, and follow scripts efficiently, but it doesn’t naturally invent weird behavior. Human testers, on the other hand, excel at creating chaos. McInerney’s view is simple: real players don’t behave like neat test cases, so testing can’t rely solely on predictable automation.
One of his favorite tactics while testing Fallout 4 was watching memory usage in real time on Xbox One hardware. With 8 GB of RAM to work with, creeping past the limit could cause instability, performance issues, or outright crashes. Rather than waiting for problems to appear naturally, he treated RAM like a “hot and cold” game, monitoring a readout and asking a deceptively direct question: how can I break this?
He noted that this wasn’t exactly standard practice. Many QA workflows focus on reproduction steps and structured checks, not deliberately trying to create memory leaks or force edge-case scenarios. McInerney decided to do exactly that, and he did it in the most Fallout way possible.
To build the ultimate stress test, he used console commands to grant himself an absurd amount of experience, rocketing his character to around level 247. Then he equipped a unique nuke launcher that fired two nukes, plus an add-on that caused each nuke to spawn ten more. With that setup, he roamed the Wasteland unleashing a storm of nuclear explosions—an extreme scenario that quickly exposed weaknesses. According to McInerney, that single morning of “super-nuking” uncovered four separate crashes.
Those crashes didn’t stay quiet, either. At the time, crash events triggered automatic reports that were emailed up the chain, meaning a lot of people at ZeniMax saw the results—reportedly including Robert Altman. McInerney joked about how his experiment essentially broadcasted his destructive testing session to leadership.
Looking back, he summed up his role with a line that captures the heart of human QA in a way no automation pitch really can: he described himself as “professionally stupid” in a way a machine couldn’t dream of. In other words, when it comes to breaking games, the most valuable skill isn’t just technical knowledge—it’s imagination, curiosity, and the willingness to do something no ordinary test script would ever suggest.
For anyone interested in game development, Fallout 4 QA testing, or how open-world games get stress-tested on console hardware, McInerney’s story is a reminder of why quality assurance remains such a uniquely human part of making games.
