Basis development is structured into "Eras." Each era adds a more sophisticated mathematical model on top of the previous one. The layers are cumulative, not replacements. Each new era builds directly on the foundations of the last.

The Question: Are these states moving together?

• Idea: State updates are discrete signals. In a clean architecture, every hook should represent a unique, independent source of truth.
• Heuristic: Temporal Lead-Lag Sampling. The engine keeps a 50-frame ring buffer for every hook and samples correlation across three temporal planes:
  1. Sync Plane ($\tau=0$): Variables update in the exact same frame (Duplicate State).
  2. Lead-Lag Planes ($\tau=\pm 1$): Variable A updates, then Variable B updates one frame later.
• What it unlocked: Basis can now distinguish between Duplicate State (variables always updating together) and Causal Sync Leaks (where an effect forces a second render).

The Question: Is this local state just a copy of Context?

• Idea: The state space $V$ should be a Direct Sum of independent subspaces: $V \approx U \oplus W$ (where $U$ is local and $W$ is global).
• Heuristic: Subspace Integrity. We instrument createContext to identify global anchors. The engine checks that $U \cap W = \lbrace 0 \rbrace$.
• What it unlocked: Detection of Context Mirroring. Basis can tell the difference between the source of truth (Context) and its shadows (redundant local hooks), and suggests deleting the redundant code.

The Question: Which bug should I fix first for maximum impact?

• Idea: Application updates form a directed topology. Importance is determined by Spectral Influence.
• Heuristic: Eigenvector Centrality. We model causal links as a Directed Adjacency Matrix and use Power Iteration to calculate the centrality of each node.
• What it unlocked: Identification of the "Prime Mover." In apps with dozens of alerts, Basis finds the specific root cause that drives the most downstream activity. Instead of a flat list of problems, you get a ranked priority list for refactoring.

The Question: Does this state carry real information, or is it noise?

• Idea: Information is the reduction of uncertainty (Shannon Entropy).
• Heuristic: Temporal Entropy Analysis $H(X)$. We measure the information gain of each state pulse. If a variable pulses at 60Hz but is near 100% predictable, its entropy is low.
• Expected outcome: Filtering architectural noise. Basis would identify "jittery state" like unthrottled scroll listeners that burn CPU cycles without contributing unique information to the system.

The Question: How many hooks does your component actually need?

• Idea: High-dimensional state often sits on a low-dimensional Manifold.
• Heuristic: Heuristic Dimensionality Reduction. Using techniques inspired by PCA, we analyze whether $N$ independent hooks are effectively operating in a much smaller $K$-dimensional space.
• Expected outcome: If 5 independent booleans always move in a rigid pattern, Basis would suggest: "Your state is effectively 1D. Replace these 5 hooks with a single useReducer to match the actual intrinsic dimension of your logic."

Each era increases the depth of what the tool can reason about. The goal is to go from detecting bugs to suggesting architecture.