"This sentence is false." Classical logic chokes on it: call it true and it is false, call it false and it is true. This playground takes a different stance, borrowed from the revision theory of truth: instead of demanding a single answer, let the truth value update in time. The Liar then stops being a contradiction and becomes an oscillation. Networks of self-referential sentences become dynamical systems with cycles and attractors, and the dynamics underneath are exact, which the calibration pins.
Each sentence has a truth value that updates from the previous step according to what it asserts:
All sentences update synchronously from the previous global state, the same discrete-time convention as a cellular automaton. The calibration checks the single-step logic of these rules directly (the biconditional, the negating assertion).
Run the Liar and it produces ...true, false, true, false..., a cycle of period 2. The calibration confirms it. The truth-teller, by contrast, is a fixed point, period 1. This is the heart of the revision-theoretic reading: a paradox is not a broken statement but a statement with no stable value, whose revision sequence never settles, instead it cycles.
Because the global state is a finite vector of bits, the trajectory must eventually repeat (pigeonhole), so every run ends in a cycle. The first repeated state marks where the attractor begins, and the distance to it is the period. The calibration verifies the detector on a constructed two-state alternation (period 2). This eventual periodicity is a theorem, not a hope.
Wire many sentences together, mutual negation, reference rings, mixed assertions, and the cycles get richer: longer periods, transients before the attractor, sensitivity to the initial assignment. Calling the resulting stable temporal patterns "morphogenesis" is a metaphor: just as morphogenesis is the emergence of biological form, here logical form (rhythm, cycle, fixed point) emerges from local self-referential rules. It is a lens on self-reference, not a model of development, and the assumptions panel says so.
Most rules look back one step, but a "percent controller" looks at a moving average of its own recent history and tries to hold its truth rate near a target. That adds memory and changes the achievable cycles. It is an honest extension; the calibration deliberately targets the memoryless rules, where the periods are unambiguous.
The dynamical reading is one of several responses to the Liar. Truth-value-gap theories say the Liar is neither true nor false; glut (paraconsistent) theories let it be both; the revision theory used here says it has no stable value and models the unstable revision process itself. The playground commits to the revision/dynamical view because it is the one that turns into something you can watch and measure.