Every lifter faces the same recurring question, is it time to add weight?, and most answer it with the two worst available tools: feel ("that felt strong") and impatience ("it's been two weeks"). Feel lies under adrenaline, and calendars don't watch you squat. There's a better referee. You film one set a week anyway: let the video vote.
The principle: load is a microphone
Weight added to a squat doesn't change the squat. It amplifies it. Whatever pattern you bring to the heavier bar gets louder: a clean pattern gets stronger cleanly, and a fraying one frays faster, with the fray landing on knees and lower back. So the progression question is never "did I complete the reps?": grinding completions is exactly how mistakes get load-bearing. The question is: did the shape survive the current weight? If yes, the next weight is earned. If no, more load just turns up a microphone on a problem.
The three-question test
Film your top working set (side-on this week if that's your rotation, front-on next) and ask the footage three questions:
1. Did depth hold, rep one through rep last? Not "did the first rep hit depth": did the last one? Depth quietly creeping up across a set is the most common early sign the current load is at your pattern's edge. Compare the bottom keyframes; the last rep tells the truth the first rep was hiding.
2. Did the torso keep its shape on every drive? Chest and hips rising together, no dip-first moment where the hips shoot and the back catches. This is the mistake that loves to appear exactly one weight-jump before you notice it.
3. Did the reps repeat like a metronome? This is the sleeper criterion, and CoachVision measures it directly: rep evenness, how consistently your reps run on one clock across the set. A high evenness number means every rep practiced the same squat and the load is genuinely owned. A fraying number (later reps rushing the descent, grinding the drive, bouncing the bottom) means the weight is being survived, not owned. Evenness is usually the first number to move when a load is one jump too early, which makes it the most useful single progression signal on the board.
Three yeses: add weight, the smallest increment available, because small jumps pass this test again next week and big jumps gamble the streak. Any no: stay, fix what the video named, and re-test. Staying at a weight until it's boring is not a plateau. It's how plateaus are prevented.
One thing you won't get from the app, on purpose: a load recommendation. The camera can't see how heavy the bar is or how close to your max you were, so it will never say "add ten pounds". It shows you whether the shape survived, and the loading call stays yours. That's the honest division of labor.
The drill: Box Squats to Depth
When question one keeps voting no, the box is the enforcer: set at your honest depth, 3 sets of 8, touch-pause-drive. The box makes depth non-negotiable at the current load until it's automatic, at which point the test passes itself.
One filmed top set a week, three questions, smallest jumps. Progression stops being a feeling and becomes a habit with receipts, and the receipts compound faster than the gambles ever did.