Version 4
SHEET 1 996 680
WIRE 208 -128 48 -128
WIRE 784 -128 208 -128
WIRE 784 -96 784 -128
WIRE 720 -48 528 -48
WIRE 208 -32 208 -128
WIRE 48 64 48 -128
WIRE 560 80 416 80
WIRE 672 80 640 80
WIRE 784 80 784 0
WIRE 784 80 736 80
WIRE 816 80 784 80
WIRE 784 160 784 80
WIRE 896 160 784 160
WIRE 272 176 256 176
WIRE 368 176 352 176
WIRE 528 176 528 -48
WIRE 784 176 784 160
WIRE 896 192 896 160
WIRE 208 224 208 48
WIRE 256 224 256 176
WIRE 256 224 208 224
WIRE 368 224 368 176
WIRE 416 224 416 80
WIRE 416 224 368 224

Strange circuit... <100k across "probe" will cause slow clicks to emit from speaker.

And likely damaging the PNP, or at least causing avalanche breakdown of it.

(By "noise generator" I mean electrical noise, not audible noise.)

I think it's a good example of a circuit designer being clever... and that's why a lot of these "novelty minimalist" circuits are pretty awful for learning electronics.

Totally agree... circuit operation depends heavily on the parasitic elements of chosen loudspeaker, and electrical properties of the "probe", whatever that is. Something as simple-looking as some inter-digitated PCB traces incorporates parasitic capacitance and inductance, which creates a dependency on the physical dimensions and properties of the PCB used.