Imagine a horse is tethered to a cart. According to Newton's third law, when the horse pulls on the cart, the cart will also pull backwards on the horse. Since the two objects are attached together, they are technically the same object, and they cannot accelerate
Let's change the situation a bit, and change the setting to - space. Replace the horse with an astronaut, and replace the cart with something heavy - maybe a toolbox of some sort. The astronaut is freefloating outside of the International Space Station, doing some maintenance work. When the astronaut pulls on the toolbox (via the tether), the toolbox will also pull on the astronaut. These two forces act on separate objects, and are directed at one another.
The astronaut and the toolbox collide.
So it's not that they cannot accelerate - these two forces don't cancel out, as they aren't acting on the same object. (Or, if you want to treat them as a composite object, keep in mind that it's not a rigid body.)
Well, the same thing happens with the horse/cart scenario, except that the horse has the ability to resist the reaction force because it is in contact with the ground. The cart itself rolls on wheels and will start to go when pulled - if we ignore friction, and vertical forces (gravity, normal force), there's a single, non-zero force acting on it. So why the cart moves is not mysterious. Of course, the horse moves as well, and the two act roughly as a unit, so we can say that both the horse and the cart have the same acceleration. For this to be true, the horse needs to produce a forward-directed force, larger in intensity than the reaction force, so that the net force on the horse provides that particular acceleration (the image below I found here shows a similar situation; it ignores friction, drag, etc.).

One point that's perhaps counter-intuitive is the fact that the forward-directed tension force doesn't have the same magnitude as the propulsion force $\vec{D}$ produced by the horse/truck.
If $\vec{F}$ is the net force on the horse/truck, note that $D = F + T = m_1a + m_2a = (m_1 + m_2)a$, so if you consider the system as a whole, $\vec{D}$ acts on the combined masses of the two objects, while internally the tension $\vec{T}$ only acts on $m_2$. (Note also that in this model the objects are treated as if connected by an idealized massless cord.)