.me

.me Kernel Phases (0–8) ​

This document defines the behavioral contract of the .me kernel, validated by tests/phases.test.js.

Note: This page documents the semantic/runtime phases 0–8. The separate performance delivery phases (storage, bounded residency, exact search, IVF) are summarized in Kernel Benchmarks.

Each phase represents a foundational capability that builds upon the previous ones, culminating in a fully expressive, secure, and observable personal semantic engine.

Phase 0 | Identity + Secret Scope ​

Core Idea: Establish identity and introduce structural privacy through stealth roots.

js
me["@"]("jabellae");
me.finance["_"]("my-secret-key-2026");
me.finance.fuel_price(24.5);

Expected:

  • me("finance") -> undefined
  • me("finance.fuel_price") -> 24.5

This phase proves that secrecy is structural, not global.

Phase 1 | Structural Selectors ([]) ​

Core Idea: The kernel treats numeric and string keys as first-class structural elements.

ts
me.fleet.trucks[1].km(1000);
me.fleet.trucks[1].fuel(200);
me.fleet.trucks[2].fuel(350);
me.fleet.trucks[2].km(1200);
me.fleet.trucks[3].km(800);
me.fleet.trucks[3].fuel(150);

Expected:

  • me("fleet.trucks[1].km") → 1000
  • me("fleet.trucks[2].fuel") → 350

This establishes that .me is a true semantic tree, not just a plain object.

Phase 2 | Broadcast Iterator [i] + Derivation = ​

Core Idea: One rule can be applied across an entire collection declaratively.

ts
me.fleet["trucks[i]"]["="]("efficiency", "km / fuel");

Expected:

  • me("fleet.trucks[1].efficiency") → 5
  • me("fleet.trucks[2].efficiency") → 1200 / 350
  • me("fleet.trucks[3].efficiency") → 800 / 150

This is the birth of mass reactive logic without loops or manual subscriptions.

Phase 3 | Logical Filters & Filtered Broadcast ​

Core Idea: You can query and mutate subsets of the tree using declarative predicates.

ts
me("fleet.trucks[efficiency < 4.5]");
me("fleet.trucks[efficiency < 4.5 || km > 1100]");
me.fleet["trucks[efficiency < 4.5]"]["="]("alert", "true");

Expected:

  • me("fleet.trucks[efficiency < 4.5]") returns only truck 2
  • me("fleet.trucks[efficiency < 4.5 || km > 1100]") still returns only truck 2
  • me("fleet.trucks[2].alert") → true
  • non-matching trucks keep alert as undefined

This phase introduces declarative selection and conditional mutation.

Phase 4 | Range & Multi-Select ​

Core Idea: Precise, deterministic slicing of indexed data.

ts
me("fleet.trucks[1..3].efficiency");
me("fleet.trucks[[1,3]].efficiency");

Supports both contiguous ranges and sparse selections.

Phase 5 | Transform Projections (Read-only) ​

Core Idea: Compute derived views without mutating the underlying tree.

ts
me("fleet.trucks[x => x.efficiency * 1.2]");

Returns a projected shape while leaving original data untouched.

Phase 6 | Cross-Scope Contract Integrity ​

Core Idea: Public and secret data can participate in the same derivations securely.

ts
me.fleet["trucks[i]"]["="]("total_cost", "fuel * finance.fuel_price");

Expected:

  • me("fleet.trucks[2].total_cost") → 350 * 24.5

The kernel correctly resolves mixed public/secret dependencies while preserving stealth.

Phase 7A | Temporal Rehydration (Memory Replay) ​

Core Idea: The entire history is portable and deterministic.

ts
import ME from "this.me";

const memories = me.inspect().memories;
const me2 = new ME();
me2.replayMemories(memories);

me2 must behave identically to the original, including secret stealth.

Phase 7B | Atomic Snapshot Rehydration ​

Core Idea: Full state portability including cryptographic planes.

ts
import ME from "this.me";

const snapshot = me.exportSnapshot();
const me3 = new ME();
me3.hydrate(snapshot);

Preserves public state, encrypted branches, secrets, and noises exactly.

Phase 8 | Incremental Intelligence + Observability ​

Core Idea: The kernel must be efficient and fully explainable.

  • Derivations register precise dependencies (ref → targets).
  • Only affected nodes recompute on change (inverted dependency index).
  • me.explain(path) provides complete traceability, including masked secret origins.
  • Invalid executable tokens remain declarative strings instead of arbitrary code execution.

Example:

ts
me.fleet["trucks[i]"]["="]("total_cost", "fuel * finance.fuel_price");
me.finance.fuel_price(30);

const trace = me.explain("fleet.trucks[2].total_cost");

The trace must show inputs, origins (public / stealth), masking, and dependency graph.

This progression is the real contract of .me:

From simple identity and privacy → to structural collections → to declarative logic at scale → to secure cross-scope computation → to full temporal and cryptographic portability → finally reaching observable, incremental intelligence.