There's an article on Table Top Builds explaining how casters aren't really squishy in 5e. Back in the day, Wizards used to have a d4 hit die. Maybe buffing casters' bulkiness was one of the causes of the issue.

I aimed to decrease casters' bulkiness and increase martials' bulkiness slightly.

With those changes, hopefully, casters can have insane utility and deal loads of damage but die easily when unprotected by martials, encouraging team play and making martials more useful.

Also, notice how Shield and other spells that give bonuses to AC don't really do much if casters take feats to be proficient in Medium or Heavy armor.  

HP & AC:

BONUS TO AC FROM SPELLS:

• Mage Armor gives a bonus of +3 to your AC. Barkskin gives a +6 bonus.
• They do not stack, similar to how unarmored bonuses work. For example, if you cast Mage Armor and then cast Shield, your magic bonus would be +5 instead of stacking and becoming +8.

ARMOR:

• Shield: Give a +2 bonus to AC that does not count against the maximum allowed by armor, but requires to be wielded in a hand

• Light Armor: Gives a bonus to AC

• Medium Armor: Give a bonus to AC but only allow a maximum +2 of other bonuses

• Heavy Armor: Give a bonus to AC but does not allow other bonuses

Making armor allow/situationally-allow/disallow other boosts to AC feels like a weird way of doing it. Why can't someone in plate benefit from three-quarters cover?

Quote from Quar1on >>

Making armor allow/situationally allow/disallow other boosts to AC feels like a weird way of doing it. Why can't someone in plate benefit from three-quarters cover?

Good point. 

You could argue that due to their armor, they lack the flexibility to get cover, so they are either in total cover or not.

Quote from CarlitosTeded >>

Quote from Quar1on >>

Making armor allow/situationally allow/disallow other boosts to AC feels like a weird way of doing it. Why can't someone in plate benefit from three-quarters cover?

Good point. 

You could argue that due to their armor, they lack the flexibility to get cover, so they are either in total cover or not.

Three-quarters cover is described as something as small as an arrow slit. If you're shooting into a castle through a wall that has an arrow slit, you'd have just as good a chance of hitting the heavily armored knight on the other side as you would if they were standing 20 feet from you in an open field?

Quote from Quar1on >

Three-quarters cover is described as something as small as an arrow slit. If you're shooting into a castle through a wall that has an arrow slit, you'd have just as good a chance of hitting the heavily armored knight on the other side as you would if they were standing 20 feet from you in an open field?

 You got me there.

We could say taking cover imposes disadvantage to attack rolls against you. Mathematically, it should do almost the same thing.

We could also say taking cover does not count against the maximum amount armor allows, similar to shields.

I will pop in and note that you are also limiting player options for design of sub-optimal or edge case play by doing this, which by most measures are more important.

Next, you are making AC a part of Class features. This is generally in direct opposition to the entire concept of AC, and ultimately would mean changing h C of all the monsters in the game in order to align with that. I will totally grant that it was like that in some previous editions, but it is not like that in 5e, and so you would have to make that adjustment across the board.

Because by tying AC to class, you are requiring that monsters be tied to class as well -- Ac is a single function across all its uses, like Hit Points are are.

And since AC is one of the core combat functions, the ripples from it are going to strike things I haven't thought on yet -- resistance, immunity, sell attacks, and more.

WHen you play around with hit points and AC like this, you impact every beign in the game.

Another example is that a Commoner -- the default, everyday, ordinary, run of the mill person -- would have to have their AC reduced to under 8, and since even Mages are supposed to be more able than Commoners to resist damage in 5e, that means you would be resetting the default AC base to 6 across the entire game -0- shifting the entire balance across the whole.

You aren't changing the way Hit Points work, so that stuff isn't affected, but you are changing the way that Armor Class works as a whole. and in trying to do so to balance it against these limited cases, you are forgetting that you have to balance it against everything in the game that has an AC.

And that means you are changing everything in the game that is set up to work around or with AC.

As an example, an unarmored Fighter in your system can now walk up to a commoner and kill them without making any rolls unless you want to see if he fumbles.

Since you mention Math well...

You break all the math for combat by doing this, because it all revolves around AC, which is the primary combat Difficulty measure. So weapon damage has to be reduced, spell damage has to be reduced, (and yes, because you changed the AC of a commoner *without a class* to 6 or 7) and the nature of bonuses to attack rolls, so proficiency bonuses, and...

You probably get the idea, think I am insane, and don't care.  Which is cool -- it isn't going to be part of my game, and they aren't going to adopt it at Wotc as the basis for everything, so it is fine for your game and I encourage you to use it for your own stuff.

How does multiclassing work within this system?

Quote from Quar1on >>

How does multiclassing work within this system?

 It doesn't.

But if you wish to multiclass, I'd say you always have the lowest AC. 

You pick the lowest starting HP, and then you count your HP based on how many additional levels you have in each class.