some sort of building blocks which are a few times larger than average molecules
I'm not sure what an "average molecule" is, but water is about a quarter of a nanometre aross. A single keratin helix is about 45nm long. In any case, you're clearly talking about nanoscale mechanical devices here, so I'll just assume you mean "nanobot" or "nanomechanism".
My fictional world has special sorts of vehicles that can shape-shift,
Obligatory grumpy question: why? What's the benefit? Non-shifting muiltipurpose vehicles can dedicate their mass, brains and energy-storage to other, more useful purposes. Or be cheaper, lighter and simpler, its all good.
That aside, how dramatic are these shape shifts? My boss has a car that shapeshifts into a car without a roof at the press of a button. It is nice at this time of year, and doesn't require ultratechnology, but turning it into something that could fly or go underwater (and come back up again) is obviously much more challenging.
The more complex your changes, the more complex your nanomechanisms need to be, and the more brains the control system must have, and the more power it must provide. Keep it simple.
The more rapid your changes, the more problems you're going to have with heat dissipation. Nanomechanisms don't shed heat easily; they're too small. Try to do too much too fast and they'll just cook themselves.
Big problem: where are these things going to get the energy they need to connect, disconnect or change shape? They certainly can't store it internally... nanoscale power storage is complex, and there's a limit to how small you can make it. You could carry around a big non-transforming power source, of course, but it might not be universally applicable to all forms your vehicle can take.
Next big problem: where do these things keep their brains? How does an individual nanobot know how to shape itself and behave in order to be part of a car? There's a bootstrapping problem here; until you've assembled some into a control system, you can't do anything complex, but they can't assemble themselves into a control system because they can't know how!
These building blocks are metals - they can magnetically attach themselves to each other, as well as being good conductors of heat and electricity
Problems: rust, demagnetisation, sticking to ferrous metals you didn't want them to stick to. The blocks obviously can't be entirely made of metal, because then you'd have real problems forming thermal or electrical insulation, and you're discarding the heat and wear resistance that various kinds of ceramic have. Also, these are vehicles... are the wheels going to be made of metal, too?
At a molecular scale, using magnetic connection seems a bit weird because, y'know, physics already provides you with a whole suite of intermolecular forces which are probably going to be easier to work with and a fair bit stronger than molecule-scale magnets. This is what living nanomechanisms use after all, so they're clearly good for something.
and these building blocks, through instructions given to them wirelessly from another terminal,
Sending radio signals to nanomechanisms is... problematic. Antennae cannot be made arbitrarily small; your nanomachines are so much smaller than the wavelengths of the signals you might want to receive that there's no way they'll ever be able to detect anything useful.
Visible light might work ok, but that means that you can only send signals to stuff of the outside, and you then need to have arranged a communication network inside your structure to propagate the control signals. When you transform your vehicle, that control network has to remain intact during transformation, otherwise you're going to have real problems ending up the right shape or long delays whilst you reform it, nanobot-by-nanobot.
There are research projects on nanoradio, describing devices which are bigger than the ones you're thinking of (and not made of metal, or magnetic), but which still have problems actually communicating any useful distance (like, from the inside of a blood vessel to the outside).
Now, I am curious about the plausibility of this - is this possible? If not, how can I make it more possible?
The world is full of self-replicating multipurpose nanomachines capable of working together co-operatively, and holding a fairly complex set of instructions for their own operation and dealing with the problems of nurient distribution and waste rejection, so the basic idea isn't totally far fetched.
Beyond that though, the practical reality of this stuff is still pretty hotly contested. At least you're not proposing that these things self replicate!
Also an optional question: exactly how big would such a building block be on average?
If you're trying to come up with something smaller than the smallest living organisms, you've probably ventured too far into the realms of implausible magic scifi. Plant and animal cells are a few tens of micrometres across. Chances are, any artificial equivalent is going to be bigger than that. Individual nanomechanisms, like real world ribosomes might be smaller, but you'll probably want to have a bunch of those per nanobot and wrap the whole thing up in a protective envelop and so on. Incidentally, if you've not already read about ribosomes you should, because those things are downright amazing.
If you're not familiar with the idea of utility fog, you should probably read up on that, too. Many people have put a lot of time into thinking about nano- and micro-scale mechanisms and intelligent systems, and I get them impression you haven't read nearly enough of what they have to say, both for and against this sort of stuff. Time spent hunting these arguments down will be well spent.
