The Algorithmic Weltanschauung (platonic spaces talk)

Consciousness, time, and reality are all facets of the same thing: compression.

The paper's central bet is that you can start from one undeniable fact — there is structured experience, right now, for you — and derive almost everything else. Not as metaphor, but as formal consequence. The move is to treat "structured experience" as the axiom and ask: what must be true for structure to exist at all? The answer is two things: a raw experiential substrate (something it is like to exist) and mathematics (the science of structure itself). These aren't two substances but two aspects of one reality, which the paper calls the Unicum. This is a form of dual-aspect monism, closer to Plato than to Descartes, and it sets the stage for everything that follows.

From mathematics as the bedrock, the paper constructs time and computation. The universe is imagined as a static logical tiling — think Wang tiles, which are known to be computationally universal. Time isn't fundamental; it emerges when you slice the tiling along a privileged direction, turning a static geometry into a sequential derivation. Crucially, the tiling is globally uncomputable (no algorithm can generate the whole thing) but locally computable (agents inside it can predict their neighborhood). This is the "Bit Bang" cosmology: a deterministic but unpredictable universe, consistent with physics as we know it. Within this tiling, an "algorithmic soup" arises, and some computational patterns persist longer than others — measured precisely via mutual algorithmic information across time.

The persistent patterns that actively maintain themselves by modeling and acting on their environment are what the paper calls algorithmic agents. An agent has three necessary parts: a Modeling Engine that compresses incoming data into world models, an Objective Function that scores model states on a valence scale, and a Planning Engine that simulates futures and picks actions. The paper proves a formal version of the Good Regulator Theorem — any system that successfully regulates its environment must share algorithmic information with it, i.e., must contain a model of it. World-tracking then forces the agent's internal dynamics onto low-dimensional manifolds whose geometry is shaped by the symmetries of the world (formalized via Lie pseudogroups). The central hypothesis is that the geometry of these manifolds just is the structure of experience — not correlated with it, but identical to it.

Subjective time gets a clean treatment: it's computational density. An agent in pain runs its planning engine frantically, generating many internal "ticks" per physical second, so time dilates. A bored adult on autopilot generates few updates, so time flies. A child learning the world from scratch is algorithmically busy all day, so childhood feels long. These aren't just analogies — the paper frames them as direct predictions of the chronoception definition. The paper also extends the framework to emotion (a tuple of model state, valence, and arousal level), to computational psychiatry (depression as a dynamical trap in the objective function), and to ethics (kindness formalized as an agent whose objective function is structurally entangled with another's simulated valence).

The scope is genuinely ambitious — from Wang tiles to the hard problem of consciousness to clinical psychiatry to moral philosophy — and the paper is honest that much remains open. But the logical skeleton is tight: one axiom, one construction, one closing of the circle.