The way the author trims the output voltage is also wrong: these small modules come by default with the trimmer feature enabled, that is, if you need a voltage different from those written near the pads just leave the pads alone and use the trimmer. But if you need one of the fixed voltages available, then before soldering the corresponding pads, the re's a thin track that needs to be cut in order to disable the trimmer so that a vibration or touching it by mistake won't affect the output (small trimmers like that one are the weakest point in every cheapo module). The thin track is just left of the "ADJ" pad; if you look closely it closes the trimmer pad to make the board adjustable by default so it has to be cut to use fixed voltages.

Also, the chip used in the regulator module is a MP2315 whose safe input voltage can be between 4.5 and 24V, which is ok for a 12V input like in the photos, just to add some data since it wasn't specified.

[1] Never, ever, ever, ever base any project on the absolute maximum ratings of any part. Especially when they're completely misleading as in this case: the 2A which the 1n4148 should sustain (it can't) are related to a peak surge of 1 microsecond, and not even close to the continuous current which is a fraction of what a Raspberry PI can draw.

The 802.3at-specced ones are a little more efficient than the af for the same power draw, I'm using them at home in a couple of applications where I'm feeding 802.3af but the power budget is cut a little close.