Thinking about competitive integrity
Essays on trust, evidence, and the technology behind verified play.
The Argument Nobody Made
A federal appeals court struck down one count against a man who used AI to generate child sexual abuse material. The judge who wrote that opinion then wrote separately to point at the argument the government never raised.
Coverage of the Seventh Circuit's Anderegg decision says a court held AI-generated child sexual abuse material is protected speech. The opinion says something narrower and more useful: one count fell, three stand, and the concurrence names the argument that might have saved the fourth. We read the opinion and the bill everyone is pointing at.
A Settlement Is Not a Verdict
A court entered a real, enforceable judgment against Meta on 26 August. Inside it, Meta admits nothing and denies everything. Both are true at once, and most of us have no word for that.
The Meta agreement is a court-entered judgment containing no finding of fact, which is a category almost nobody has a word for. This is what the filing itself says, including the parts that flatter Meta and the parts that do not, with clause numbers on the contested ones so you can check us.
Things You Know and Cannot Show
A fire lieutenant pulled his crew out of a house seconds before the floor collapsed and could not say why. What his intuition actually told him is not what most people think, and it changes what yours is telling you about that guy in your lobby.
Trained instinct is real. It is also worse at one specific job than people assume: it detects that something is wrong far better than it identifies what. That gap is the whole difference between a good read and an accusation.
The look-alike problem
Your RGB lights, your overlay, and your macro pad all do things a cheat does. Here is why honest software keeps tripping anti-cheat, and what actually helps when a flag lands on the wrong person.
The software that syncs your keyboard lighting injects code into running programs. Your overlay draws inside the game's own frames. Your monitoring tool reads other programs' memory. Every one of those is a technique cheats use, which is why legitimate setups keep tripping detection built to spot the technique. This is the plain-words map of the look-alike problem, and the practical part nobody leads with: what an appeal that works actually contains.
Where cheats hide now
Anti-cheat runs at the deepest level of your PC. So how is anyone still cheating? Because the newest cheats aren't on your PC at all. Here is the plain-words map, and what it costs the honest player.
The best cheats have climbed out of reach of the software built to catch them. They moved from the game's memory, to a second piece of hardware, to a whole separate computer that just watches your screen. Once you can see that ladder, you understand why anti-cheat stopped looking for cheats and started judging your play, and why that lands hardest on the people who never cheated.
A signature is a receipt, not a verdict
Our process collector is unsigned. Here is why that is not the contradiction it looks like, and how to check us anyway.
Every dangerous driver in our Field Guide has one thing in common. They are all signed. The certificate did not make them safe. So when we shipped our own collector, we had to decide what we actually wanted you to trust. We did not buy the checkmark. Here is what we did instead.
Present Is Not Proof
We built a check that can spot genuinely dangerous things on a PC. The first thing it taught us was when not to raise the alarm.
A vulnerable driver showed up on one of our own machines, from a tool millions of gamers run. The easy move was a red flag. We didn't make it. Here is why, and what it means if you have ever been good enough to get accused.
272%
AI-driven cheats tripled this season. The industry is building bigger walls. We think that's the wrong question.
AI-based cheat detections surged 272% in a single competitive season. The detect-and-ban model is breaking. The question isn't how to detect better. It's what trust looks like when detection alone isn't enough.