Other peoples mistakes can be really instructive. I don't want to make a video showing everyone the wrong way to practice, but I don't have to because this guy Michael Pekor did it for me.
He has two major problems:
1. When he hits the padded pole sometimes part of the shock of that hit goes back into his body. You can see this because his head shakes. This is especially obvious around 1:00. Doing this for even a short time can damage your heart because force issued through the arms leads straight back into the heart. He should back off.
2. His shoulders are leaving the dantian. Again this is most obvious around 1:00 in the video. The shoulder comes forward and out. In this "out" position, power and connection are lost in every direction except the one he is hitting in. I'm not saying it won't hurt to get hit by this guy, it most certainly will hurt. Taijiquan is not just about power and fighting is also about perfecting ones art, and "shoulders outside the dantian" is a flaw.
It is imperative that if he chooses to experiment with my corrections that he fix the shock problem first! If he were to fix the shoulder problem first the shock problem would become dramatically more dangerous! The shoulder problem is currnetly functioning like a trip-switch, defusing much of the power travelling back toward the heart.
Fix the shock problem by getting rid of the padding and just hit the solid pole. If even the slightest bit of shock travels back towards your torso, your structure is wrong. Don't hit very hard or you will break your hand.
Fixing the "shoulders outside the dantian" is more difficult. The arms must first be completely empty-- drained of tension, intension, form, and conception. From that state they can gently be drawn into the dantian. Once the whole body is unified with the dantian, jing and qi are much easier to distinguish. Thus it is a simple progression from there to allowing the mind (Yi) to lead the annimated part of us (Qi) which will, without any conscious act, take the whole body mass (Jing) for a ride.