I'll start with the thing I don't want misread. Nothing on this site replaces putting on the gear and doing the job. A drill on the floor with a real line, a real ladder, a real crew, is worth more than anything I can build in a browser, and it always will be. If you take one thing from this post, take that.
The problem is what happens between drills. A company gets maybe a few real hours of hands-on a month once you subtract the calls, the inspections, the truck checks and the paperwork. A new officer might make one real fireground decision a month, sometimes fewer. That's not enough reps to get good at anything. Firefighting used to give you the reps for free; you learned by going to fires. There are fewer fires now, and the ones we get are worse. So the reps have to come from somewhere else, and "somewhere else" has mostly meant PowerPoint.
That's the gap Cognition is for. Not the drill. The forty minutes at the kitchen table after the drill, or the night shift when nothing's going on, or the day the training officer can't get the whole department in one room. You take the seat, the call comes in, you make the decisions you'd have to make, and the after-action tells you what it saw. Then you argue with it. The app is built to be argued with; there's a button for it.
The instructor problem
Here's the other reason, and I want to say it carefully, because I've done this myself.
There's a human trait in almost everyone who teaches: we want to correct people. We want to be the one who shows the right answer. It feels like teaching, and most of the time it is. But it has a failure mode. A scenario gets set up so the student has to fail, because the failure is the setup for the lesson. Push that far enough and the instructor, reaching for the correction, can end up saying something that isn't quite right, and not know it. Nobody in the room is in a position to push back, because he's the one who does the correcting. The wrong thing goes out with the authority of the senior guy behind it, and it gets written down.
That's not a knock on instructors. I've been that guy, and I'd guess most people who teach have. It's an argument for more reps that don't go through any one person's opinion. A tool that puts the decision in front of you, shows what it graded and why, and cites the SOP it graded against is something you can check. If the tool is wrong, you can see it and say so, and it gets fixed. That's harder to do with a person, including me.
What it is and isn't
Cognition doesn't know your first-due. It doesn't know what your third-alarm assignment really looks like at 3 a.m. with two guys out. It runs on a set of scenarios and a set of rules, and where the rules come from is written down inside it. When a number isn't sourced, it says so. When you disagree with a decision it preferred, you can take the whole thing, options, grading, conditions, the SOP it cited, and put it in front of your crew or your chief or whatever AI you talk to and fight it out. I'd rather it be argued with than trusted.
The people testing it now are firefighters and officers I know. The scenarios started as calls from one department. What gets built next is whatever they tell me is wrong or missing. That's the whole development plan, and I'm fine with it.
So: go drill. Pull the line. Throw the ladder. Then, when the tools are put away and the day's not over, give the seat a few more reps. That's what it's for.