Starting with your cylinders around arbitrary minor axes, I noticed that towards the beginning you were keeping your side edges parallel (basically artificially forcing your vanishing point to infinity as discussed back in Lesson 1) - fortunately this is something you stopped doing soon after, though I do feel it would be helpful for me to explain why that would be incorrect.

Basically, vanishing points only go to infinity when the set of lines they govern run perpendicular to the viewer's angle of sight, where they wouldn't be slanting towards or away from the viewer, and wouldn't be moving at all through the depth of the scene. Since we're rotating our cylinders freely and randomly through space, this is rare enough that we may as well always ensure that our side edges are converging towards a shared vanishing point, allowing for a shift in scale from one end of the cylinder to the other (so the far end ellipse ends up smaller than the end closer to the viewer).

Anyway, as I mentioned, this is something you corrected pretty quickly, so there's no sense dwelling on it further. Moving forward, you've done a good job of identifying the true alignment of your ellipses as you progressed. The ellipses themselves tend to be drawn pretty well when you allow your cylinders to take up more space on the page - but when you draw them smaller, you have a tendency to cramp up, resulting in clumsier linework and more uneven ellipses. Always take advantage of the space available to you on the page - this will help you engage your brain's spatial reasoning skill, and will also make it easier for you to remember to draw from your shoulder rather than your shoulder or your wrist.

That 'scale shift' I mentioned earlier, which comes from the convergence of the side edges, works in tandem with the shift in degree (where the far end gets wider) to convey foreshortening to the viewer - basically telling them how much of this cylinder is moving through the depth in the scene. When drawing the far end of your ellipse, be sure to consider how much to shift its degree based on how much the scale has shifted. You want them to be roughly similar, and avoid situations like 124 on this page, where the scale shift is quite dramatic but the degree shift is more minimal.

Moving onto your cylinders in boxes, here we end up circling back to that issue with your lines all being way too parallel - it looks like you put all your vanishing points at infinity there, perhaps in an attempt to simplify the problem in front of you. Unfortunately, it doesn't work that way - the vanishing points are determined by the orientation of the form. Even if that form is perfectly aligned such that two sets of its edges runs perpendicular to the viewer's angle of sight, that last one is always going to end up having a concrete vanishing point causing its lines to converge (like in any classic 1 point perspective arrangement). Of course, again, for this exercise we're rotating the forms randomly, so you really shouldn't be putting any of the vanishing points at infinity. Each set of lines should be converging, even if only very gradually.

It's worth mentioning that I do see more conscious convergences later in the set. It's maybe up to around 190 that you were keeping everything very parallel, then started to fix things after that point.

Now it does seem that you're applying the line extensions correctly for the most part - though I would strongly recommend drawing them so they actually go through the ellipses rather than starting on one end, just to make sure you can see its relationship with the whole ellipse even after you've put your ruler away. Also, drawing these boxes and cylinders larger, as mentioned before, will help a great deal.

All in all, I think you've moved in the right direction throughout both of these exercises, but you do have a number of things to keep in mind, and as far as markmaking goes, there's still room for improvement. Be sure to continue practicing this exercise as part of your regular warmup routine, but I will go ahead and mark this challenge as complete.