Left click and drag moves the stage. Scroll changes magnification. Right click records the particle centred in the box. Keys 1 to 4 set the dwell time. WASD and Q/E work too.
Start wide, find a support particle, then magnify until the small particles on it resolve. Centre one and record it. Each particle counts once, and the score rewards centring, magnification and a slower scan.
The image builds as a raster. The beam scans line by line and the current line is marked, so the frame is always part new and part old. Move the stage and the scan restarts from the top.
Contrast follows thickness through a sphere, so particles are brightest at their centre, and the small particles on top of a support are far brighter than the support itself.
Pixel size limits what exists. Particles under about one pixel are not rendered at all, which is why you have to commit to a support and magnify before you know whether anything is there.
Dwell time trades speed for noise. Noise falls as the square root of dwell time on top of a shot-noise term. In a real microscope that trade is also a dose budget, and the specimen pays for it.
The specimen never ends. Particle positions come from a hash of the coordinates, so you can drive the stage forever, and returning to a spot shows the same particles rather than new ones.