Here's an idea I'm trying to flesh out, and was hoping someone has heard of something similar. 

Several times I've built well-insulated slab on grade unheated structures.  The slab itself is uninsulated underneath.  The temperature inside these structures has never dropped below 46F.  Apparently these are geothermally heated.   If you built an attached garage this way it would only cost a few hundred dollars more than standard construction.

What if you mounted a through the wall PTAC heat pump in the wall between the house and garage?  They only cost $500, and put out 12kbtu/hr with a COP of 3.  They operate down to 24F.  I doubt the garage would ever get down to 24F, so the resistance heat would never come on.

If the house were new,small and superinsulated, one of these units is all you need.  No offense to all you great geo contractors out there, but it wouldn't break my heart if I didn't see you on my jobsite.  I just plug it in.

Remember, a smallish superinsulated house doesn't need distribution ductwork.  Of course, you'd have to seal up the fresh air damper and make it tamperproof.

Any one ever hear of this?

I built a workshop in NC with an uninsulated slab , very airtight , with insulation down the sides of the slab. The coldest it ever got with no heat was 40 degrees when it was 12 degrees outside. The 55 degree ground keeps it warm. The problem with heating a structure like this is that the 55 degree ground will suck heat out just as well. I would insulate under the slab if I was going to heat the space.

Betcha that little superinsulated shed would get MUCH colder than 46 degrees with a ton of refrigeration applied to it all night - You'd have an excellent walk-in-cooler, possibly a freezer...

A PTAC would shift to resistance mode before it happened, but you might even freeze the ground beneath the slab.

Not a shed, a 22' x 28' garage.  I don't expect it to work during design conditions, but down to 15F ambient would be great.  The PTAC shuts off at 24F, so your car is always warmer than 24F.  I wouldn't be worried about freezing the slab because the soil under it will always have less than 15% moisture as long as the site work is done right.

The next day, if ambient goes up to 40F, you could easily "heat" the garage with ambient air.   Heat loss of the house at 15F is only about 8,000 btuh

I'll follow up with a few calcs.

What I'm proposing here is a geothermal system at about 20% of the price of the status quo ground coupled heat pump system, so it's worth investigating.

My rough calcs are showing 6,000 btuh could transfer into the garage from the slab.   Almost enough to make it worth a try.  My data isn't that good.

Opus,  please give me some more specifications on your workshop, and I can make a better calculation.  What is the footprint, R values of the walls and ceiling, height of walls, and the size of the windows.  I assume they're standard double pane.

That is a very interesting idea Kevin.
My workshop is 32' x 48' with 12' high walls. There are 4 windows that face SW. They are 4'x6' each. The shop is in a valley and only gets sun from 10 to 4 in the winter. The whole building is light gage steel- 6" studs with a one inch thick piece of styrofoam glued to the inside of each stud and each truss to provide a complete thermal break on the inside of the structure. The outside sheathing is Hardipanel. The wall cavities are thus 7" and are filled with blown in fibergalass. The attic is also blown in fiberglass at R 40. The R value of the fiberglass at 7" would be about R 28 but becaue of the steel studs I would guess the whole wall R value would be closer to R 15. There are also 12 windows that are 2' x2' and one 6' sliding glass door. Everything double pane. There is a 10' x 10'6" insulated garage door on the NE end. I did airtight drywall construction and sealed all electrical boxes so the structure is very airtight except for maybe around the garage door. I did the 1" foam on the inside after talking to one of the engineers at the Oak Ridge National Lab on how to provide a thermal braek for the steel. Foam on the outside of the steel might be better but with steel trusses I would have the problem of them conducting heat out of the structure like a large raditor if they were not thermally broken on the interior. I have an insulated bubble over a large thermal mass [ the ground] with a constant temp about 55 degrees. I never have to worry about freezing pipes in the winter. In the summer it is always 10 degrees cooler inside than outside and would probably be much cooler if the large windows were shaded. It is very easy to heat in the winter if desired. I have a 35,000 BTU propane direct vent wall heater the heats the place up in no time if desired. The floor temp is often be below the dew point so if you open the window on a cool night the floor gets wet. The same thing might work well for a heated house since you would only have a delta T of about 15 degrees [55 degrees to 70 degrees] . I was very happy with the outcome. I imagine you could pump heat out of the space on cold nights annd warm it during the day with 40 degree air as you suggest. Good luck.

The heat loss on Opus's garage at 12 degrees outside and 40F inside is about 8600btuh free geothermal from the slab.

The heat flow through  the slab is operating at a temperature difference of (55F-40F)=15F.   If we can safely assume that the slab heat flow is linear with temperature, then how much does that heat flow increase at a lower indoor temperature?   Well if we can take the indoor air temperature of the shop down to 25F, now the temp. diff. is 30F.  Therefore the slab heat flow would double to 17,200 btuh.

So, a rule of thumb, at least for the soil conditions at Opus's site, is that a perimeter insulated slab can produce 11.1 Btuh/ft2 of low temperature geothermal heat for an air to air heat pump.

Next question, does code allow an unvented attached garage? 

MI mechanical code requires ventilation in living space. Not a building inspector, but I believe it would be required in a seperated attic over a garage, but not a garage. That said, do not forget a couple things;
the garages/workshops you are referring to are not actively cooled (as engineer alluded to) as they will be if you transfer heat from them. Further anything you use to insulate these areas with will impede gain just as surely as loss, and a wall (garage partition) that shows less loss on heat load (due to intermediate space) now adds additional load to building.
I'm not suggesting the idea is completely without merit, in fact it's already done (office cooling with window air conditioners discharging heat into freely ventilated factories).
Unfortunately, it's always about footprint... Will your garage be a large enough heat sink to condition the living space (95%) of the time? Will you be happy with a garage that is 25* instead of 45?
As always, there is no free lunch, so make sure you like the price/flavor.
Joe