I posted this here but I feel it warrants it's own thread. Part of what's wrong with crafting is that a lot of the best equipment is found, and the best crafted stuff is heavily subject to RNG in terms of both finding the recipe and the materials. One way to fix this is to add some sort of reinforcement system, where you can use your materials to permanently enhance an item. This would not be done randomly, like anvils, but would have a specific recipe for a specific stat boost (e.g. 2 steel ingots + any weapon = +1 piercing damage; hyperborean potion + sapphire + any armor = +1 hyperborean resistance). Recipes for this should mostly be found, but it would be nice if 1 or 2 basic recipes were known at level 1 of the skill. Like some current recipes, multiple crafting skills could have access to the same reinforcement through different materials, or each crafting skill could specialize in certain stat boosts or equipment types. Either system could be made to work. In order to balance this, there should be limits on the number of times an item can be upgraded, as well as the quality of the upgrades. Possible balance methods include max. of three +1 upgrades per item (each type of upgrade available only once per item?), or one upgrade per item, with different recipes for +1/+2/+3 (or scaling with ranks 1/3/5 of the skill). These numbers can be tweaked depending on which upgrades are available to which crafting skills, of course.
There are some similarities between our ideas, but the key difference is that mine doesn't add a new type of item to clutter up the already limited bag space. Yours is a good idea, but mine accomplishes more or less the same thing by fixing a lackluster set of skills and adding some use to materials that are hardly worth keeping (yet still being found) past floor 3. The point is that some sort of manual and permanent item customization would be spiffy, regardless of how it's actually integrated.
We've been kicking around such ideas since mid-beta, actually. We just didn't have time to actually implement it in any useful way, so let's just say that it's exceedingly likely that we'll do something with this.
