This is an astable multivibrator circuit. Assuming that TR2 is initially on, the right pin of C2 is grounded and since C2 isn't charged, the left pin of it is also at 0V. Now, Out 2 is low.
C2 is then charged up slowly through R3, until a critical moment when the left pin of C2 passes around 0.7V to activate TR1, in which the circuit flips and TR1 turns on, causing Out 1 to be low, and the cycle repeats.
However this has a problem with the rise time. Assuming that TR2 is initially on, TR1 is off and thus Out 1 will be high. However, the capacitor will cause the rising edge to be a curve and not "instant".
This is an improved astable multivibrator circuit, which improves the rise time by using a diode to prevent the capacitive action when TR1 is high (assuming TR2 is initially on). Instead, C1 will be charged through a separate resistor.
It works in real life, but I just can't wrap my mind around it. If D2 is there, and TR2 is on causing the right pin of D2 to be grounded, the left pin of D2 will be around 0.7V (forward voltage drop of D2). As C2 is not charged yet, the right and left pin of it will be 0.7V.
But doesn't this cause it to achieve the critical level of 0.7V already? This is where my head explodes from infinite frequency.
How does it work?
