The main issue I see with this is the vast amount of combinations. If each combination and order yields a unique item with a unique orange text, Blizzard would basically have to have an algorithm generating icons/skins/orange text affixes. You can't just sit down and design millions (or more) of new items for obvious reasons. So, that is out the window. This leaves us with duplicates. To put it another way, you'd need to have, for example, [[Furnace + Sunkeeper] + Ring] and [[Furnace + Azurewrath] + Ring] yield the exact same ring. And then numerous other combinations are going to yield that same ring. Let's call this the Ring of the Vizjerei (RotV). If I am misunderstanding and this recipe would not always yield RotV and instead yield a random legendary ring, then I guess I am missing the point of having different combinations and the order of said combinations being meaningful. The way you describe it in the video, my understanding is that each combination is a recipe, of sorts.

So, let's say I want to synthesize a RotV with 50% crit damage. I'm always using a ring as my final reagent, so I can just pull that 50% CD through from that parent ring. Because there will be redundant recipes, it really doesn't matter if I used [Wormwood + String of Ears + BK Wedding Band] or [Furnace + Sunkeeper + BK Wedding Band]. Remember, you can't possibly have each combination give a unique result. But as long as I have that 50% CD BK ring at the end and some combination of reagents that yield RotV, I can get my 50% CD RotV. If I understand correctly, I could also get that 50% CD from one of the earlier reagents, let's say one recipe for RotV uses Gloves of Worship, then I can use any BK ring and pull my 50% CD off the gloves.

What if you used fixed recipes instead? Let's say Blizz creates 10 (or 20, or 5, or whatever) new items of each slot. Each one has a set recipe with some flexibility. In the last paragraph, I used [Wormwood + String of Ears + BK Wedding Band] as a possible combination for RotV. Let's reduce the specificity of the items down and add an additional ingredient to give us the formula: [Any Staff + Any Belt + Any Glove + BK Wedding Band] = RotV. You could even allow the desired stat to be carried over when the value is too high to roll naturally on the resultant item. i.e. If we want to carry intelligence to our RotV, we could pull the roll from any of the items but if you pull it from the staff or the glove, it could be reduced to be within the appropriate roll range of a ring. e.g. If our glove has a max int roll (750), the crafting window will indicate that our mystery ring (actually a RotV) is going to have 500 int (or max ancient int roll) on it. If the glove only has 736 int, our UI will inform us that the product is going to roll with 490 int, with a chance to roll as a top end ancient roll, should we get an ancient RotV. Perhaps one Sword recipe is [Any Bracer + Any Amulet + Exarian]. This could have a weaker legendary power as you could churn them out with ease. Then maybe we've got a sword recipe that is [Any Ring + Any Boots + Any Glove + Any Shield + Shard of Hate]. This item could be significantly stronger. Maybe cap this system at 5 total reagents to avoid confusion.

Moving on the the second system. What you essentially have here are two different systems. The "anything goes" system and the "combine 6 items of the same type" system. The combine 6 items system is based on carrying over a legendary property. Only have one recipe (product) per item type. e.g. Six Wizard Hats always yields a hat called Jaetch's Ninja Mask (JNM). Because our goal here is to bring one legendary property from the original six, the additional legendary property on JNM can't be too OP by itself. Our result gets four primaries, JNM's natural orange text, an orange text from a reagent or a blue text secondary from one of the six hats used. Let's say we want the Swami property on our JNM. This is only giving us a 1/12 chance of getting the affix we want. But we can do it every time we find a Swami and five other hats. One could conceivable craft a Swami'd JNM pretty often. A dedicated player could craft a good (i.e. good primaries) Swami'd JNM every couple months, perhaps.

My final thought is that I think this would have a negative impact on stash space. People are going to be saving up all these items to use to make, say, a Bracer with the Strongarm property. A way to alleviate this could be to store the reagents at the mystic. For those of you who have a seasonal char that has been rolled over to NS on the PTR, you know your message box is full of all the items that were in your stash. It's basically an alphabetical list categorized into "legendary," "set," "consumables," etc. (we wouldn't have those categories since we are only dealing with legendaries) I propose you take an item that you know you want to use as a reagent to Myriam and irrevocably add it to her bank of reagents. This destroys the physical item, it can no longer by equipped or salvaged and, most importantly, removes the item's unique ID from the database since it's now just text strings of the affixes in the bank. It now appears in an alphabetical list (that can also be sorted by gear slot) and it can be added to any applicable recipe, at which point it is removed from the bank. You can add items straight from your inventory to a recipe or pull them from Myriam's bank. If you realize you added an item you don't want in your bank, like if you already have a gg RotV, you would be able to discard an unneeded staff. The bank would have finite space of perhaps 200 items. The interface warned you that you can never get the item back when you added it to Myriam's bank, so no one can complain that your only option is to destroy it forever. Maybe it can refund a forgotten soul just for the QQers.