The book's central concept is one that I desperately want to make work. I believe in its vision, 100%, no reservations. And the mechanical framework that it chooses to execute that vision is inspired - allow players to play creatures from the Monster Manual by making up a custom class that doles out the monster's special abilities over the course of its level progression, thus ensuring that a level 15 djinn (or what have you) is balanced against a level 15 Fighter and a level 15 Wizard.
And I think you may now be seeing where the problem comes from. This product comes late enough in 3rd edition's life cycle that they know to wallpaper over the cracks, but not late enough to actually benefit from the design experience that would lead to Book of Nine Swords or 4th edition. So you can simultaneously have a prestige class chapter that says, "prestige classes aimed at casters should be particularly attractive" (because losing caster levels outweighs nearly any ability imaginable, especially at high levels) and an Illithid class description that suggests they would make good wizards because of their Intelligence bonus, but which doesn't actually let you multiclass into Wizard until level 15 (giving you a really high save DC on your 5-6 1st level spells per day).
Funnily enough, this means that the strongest classes in this book are the ones geared towards eventually multiclassing into a Fighter, because Fighter abilities don't scale, so you get just as much taking your first Fighter level at level 10 as you would taking your 10th Fighter level at level 10.
Assuming, of course, that the monster levels 1-9 were equivalent to a Fighter in power, which is honestly pretty hit or miss. The giant classes were probably fairly close, with their high bonuses to Strength compensating for a lagging Base Attack Bonus, but even they'd cheat you out of hit points, skills, feats, and savings throws.
Which brings us, finally, to the book's fatal flaw. The thing that is the essential and primordial cause of its incurable balance issues - its bizarre fixation on allowing (forcing?) you to recreate the exact stat blocks from the Monster Manual. So many of the dead levels in this book are caused by weird and unnecessary stat increases. Because monster Ability adjustments are all based on the assumption that the Monster Manual stats came from a PC who rolled straight 10s, so if you're playing an Azer (fire dwarf, basically), you wind up a level 6 character with 2 hit dice because you need an extra +2 Str, +2 Dex, +2 Int, and +2 Wis, and none of that would be necessary if you just assumed that the NPC Azer possessed a Standard Ability Array. You don't play an Azer because you want to be slightly smarter and wiser than an average human (and a dexterity bonus doesn't fit into the archetype at all). You play an Azer because you want to be a dwarf who is constantly on fire. Ditch the superfluous stat increases, rein in the natural armor to a reasonable level, get rid of the :snicker: 13 spell resistance, and get this shit done in 3 levels, with 3 hit dice, so we don't actually have to break out the abacus to calculate this guy's feats.
This whole thing could have been an absolutely amazing supplement with just a little bit of fudging. There is absolutely nothing about the grig's power set that requires they have half a hit die at level 4. Stagger the spell-like abilities a little better, ditch most, if not all of the Ability increases, and nobody at all is going to complain about this guy having 4 levels of Bard hit point, attack, save, and skill progression, even if the MM Grig is half-HD creature.
Although, honestly, the progression of spell-like abilities is another case of the book writing itself into a corner by trying to emulate the Monster Manual. Some of the 20-level classes can sort of work as a wizard stand-in, because 12d8 is relatively close to 20d4 and +12 BAB at level 20 is close to the Wizard's +10, but the need to match the exact list with the exact frequency, winds up leading to some very eccentric builds. An efreet gaining Wish one time per day at level 18 puts it in line with a sorcerer, as do its 1/day 4th level spells at level 8 and its 1/day 6th level spell at level 12, but it misses out on the rest of the power curve in exchange for some physical abilities that would be impressive if they were not diluted by being spread across 19 fractional hit dice.
And the Efreet was one of the best classes in the book. In some cases, you have attack, durability, and spellcasting all lag behind the sorcerer's progression, all for the sake of wedging in things that players aren't really going to want, need, or even necessarily notice. The Ettercap gets a +3 BAB at level 9, for crying out loud. You'd have to really work to get a character with such low accuracy (it's not actually possible without multiclassing three times into classes with a slow BAB progression) and what you get is more webs than you can reasonably use in a day (eight) and an admittedly nasty poison that is still somehow not as effective as casting polymorph other once.
My quick fix for this is to just strip out all spell-like abilities and replace them with levels of sorcerer casting, with your known spell slots being filled with the monster's spell-like abilities, that way a creature like the green hag, who wants to be a spellcaster after her monster level progression, can slip into a level appropriate role, lagging behind only a little thanks to what basically amounts to a Fighter multiclass. And for a lot of creatures, you wouldn't even need to prefill the slots, as only a few of their spell-like abilities are actually iconic. As for creatures that gain single orphan abilities, but no real caster progression, just fudge it with a feat. Prerequisites: Monster Type, 2x spell level HD. Benefit: cast this spell 1/day as a spell-like ability. Not necessarily "balanced" per se (mainly against the crap that high level martials would otherwise get), but good enough.
Because what this book really needs is more design. A lot of these monster levels are filled with cruft that is only tangentially related to the reasons people want to play the monster in the first place (the giants' rock-throwing is barely an ability, it's more a natural consequence of being huge and having high strength, combined with early AD&D's first draft attempt at modelling that one passage from The Hobbit where giants throw rocks at each other.) They really should have focused on stripping down the monsters to the most basic reason people would want to play them (i.e. a Medusa turns people to stone, a grig is a cute little guy with a cute little fiddle, a hound archon is a furry paladin), give us just enough levels to get to that concept, and then let regular class levels fill in the rest, because the PCs will be playing unusual monsters, just like they play unusual demihumans.
Ooh, I got deeper into the mechanical weeds than I really wanted to, but in my defense, this is a very mechanical book. On the flavor side, it's got a less-severe case of the malaise that affected 2nd edition's Complete Book of Humanoids, where it's never entirely clear what a "monster" is actually supposed to be, so it kind of just treats the whole gamut as one interchangeable mass and can sometimes get problematic as a result (goblins are, at one point, referred to as "a lesser race"). However, while it fails to imagine some of the more appealing niches for monster PCs, it does at least treat them as viable characters, and that's something.
Overall, I can't say I recommend this book. I really, really want to play a game that realizes its potential, but in order to do that, I'll pretty much have to rewrite it from the ground up.
Ukss Contribution: The "gelatinous creature" template (one of 17 available in this book, more than doubling the number of published templates) use a grizzly bear as its example. Now, I can't be certain that whoever wrote that template was deliberately trying to sneak a giant gummy bear encounter into the game, I can be certain that when I put it in Ukss, I will be.