WALK. Researchers climb a hidden landscape of questions. Crowded ground attracts settlers and repels explorers, so the community can spread organically. They mostly stay within reach of existing work, but sometimes start a great leap, which may land in emptiness and found a colony. When the questions somewhere are spent, researchers migrate to where the funding is. The landscape idea is borrowed from Weisberg and Muldoon's epistemic-landscape models, which describe how a division of cognitive labour emerges among mavericks and followers.

CONSOLIDATE. An empty cell with enough worked neighbours becomes worked; interiors deepen on their own. This reflects how simple reasoning (with some Kuhnian normal science) can often fill in these gaps straightforwardly.

COHERE. Every stroke is drawn toward its neighbours' direction, weighted by their success, to model the fact that imitation follows results. Local coherence is how a pile of results becomes what Ludwik Fleck called a thought collective, with its own trained way of seeing.

EXHAUST. Work locally depletes the questions that attract it. Each living paradigm has a soft carrying capacity: past a certain size its outward growth fades and the same effort turns inward, deepening detail rather than claiming space. But exhaustion here is local and temporary. New questions keep arising wherever people work, syntheses and anomalies can create whole new landscapes, and the world grows without end. Ideas are living things that reproduce and hybridise.

SHIFT. Novelty accumulating in sparse territory nucleates a wave that re-disorders what it crosses: orientations are rewritten wholesale, though accumulated depth partly survives, mirroring how instruments and data often outlive the interpretation.

SYNTHESIZE. Where separate living bodies of work grow into contact, new questions erupt at the seam. Further combinations can result in greater growth, as Uzzi's work on recombination suggests. Sometimes the meeting founds a child field, born apart, carrying a blend of both parents' grain, much like neuroscience and computer science begetting AI.

CRYSTALLIZE. A mature, coherent body of work eventually sets: over a visible ritual its strokes anneal, then it freezes in place as part of the record while living work continues around it. Newborn work beside a crystal inherits its grain, much like textbooks teach the next generation basic theory and taste. When information reaches the world model's limits, a new era comes in and the whole world renormalizes, contracting about the fused record so that last era's entire architecture becomes this era's texture. This reflects the fact that paradigms compress knowledge, and draws on Kenneth Wilson's renormalization group.

Each patch of ground can be most productively worked via a particular approach (i.e. stroke or grain). But much like we don't know where to find discoveries, we don't know what this best stroke is. Motion could be studied via Zeno, Plato, Newton, or Schrödinger; the ground was the same, but these different approaches were vastly different in terms of their fruitfulness. Several approaches can be somewhat right at once, which is why rival schools are sometimes both correct — and why a community can herd, quite rationally, onto a method that results in 0.3 while the 0.9 sits undiscovered beneath their feet.

Depth accrues only where the method fits. Researchers cannot read the field; they feel it through what deepens, and they imitate success. Variation is introduced from outsiders and revolutions, selection by a nature that answers only some approaches, retention by textbook and tradition. Each era, part of the truth refreshes — some questions carry across a revolution, and some don't. But nothing is ever finally settled. Waves that overlap old dogma can sometimes rewrite it; easily at the periphery, more rarely toward the core, much like Quine's web of belief.

Every constant in this simulation represents an institutional choice:

• Funding sets how many researchers the world sustains.
• Agenda decides whether researchers follow the crowd or spread across communities.
• Walls decide how easily fields can meet and recombine.
• Consensus reflects how closely new research needs to match surrounding grain.
• Canon decides how final past work is.

Replay reruns the identical seed — same nature, same starting people — under your current rules, providing a sort of counterfactual history, with the previous run left on the knowledge chart as a ghost.

Clearly, this model leaves out elements of real-world science. The sliders are extreme abstractions over the good, the bad, and ugly world of real scientific institutions, built more for beauty than for precision. But we are hoping that this makes a statement of sorts: that science can be studied like a wild and precious model organism, its institutions varied like genes, and that perhaps Golden Ages await us if we do.