Frankly speaking, the way Q2 and Q3 are connected on the 2nd circuit creates the negative feedback and makes the circuit stable. The more current flows through Q2 collector, the less current flows into Q3 base, preventing further Q2 collector current increase and restoring the balance. This makes the output circuitry stable. The reference point is the junction of Q1 and Q2 emitters.
Now, let's imagine that initial voltage at Q1 base is at 0V. Q1 emitter will be at -0,6V, same as Q2 emitter. This will cause certain current flow through both transistors. At the same time, base of Q3 will be driven by some current (depending all the resistors), polarizing base of Q2. Any fluctuation of currents between Q2 and Q3 will trigger the above mentioned mechanism of self-balancing, making base of Q2 and output to stay at 0Vin order to keep the balance. Now, any change of Q1 base voltage will cause the change of Q2 base - emitter voltage, resulting in Q2 collector current change (opposite to Q1), which will cause Q3 to restore balance at the level defined by voltage at Q1 base... So the output voltage will FOLLOW the voltage at Q1 base. The common emitter resitor of Q1 and Q2 forces them to react to each other's parameter change (voltage, current, temperature).
Excuse my language, which is not native.