A1 -- Agent Genetics

note

Current public-DAV authority boundary — 2026-07-12. Pre-launch target design; nothing here proves a live system. A public-DAV consequence may occur only when at least two natural-person councilors bind the exact consequence in a complete valid bound PRISM decision receipt. PRISM records and verifies that receipt only; it never serves as council, signatory, authority, or receipt producer. AI and caste seats stage unsigned proposals only; they never authorize or execute a public-DAV consequence. Constitution or membership adoption establishes constitution and membership only; it does not authorize a later consequence. Policy may constrain an unsigned proposal but never authorizes execution or substitutes for the complete consequence-bound receipt. Before receipt validity, consequence fails closed to read-only proposal, simulation, or deterministic sandbox; live evidence remains gated_pending_complete_valid_bound_receipt. Software deterministically carries out only the exact consequence bound to a complete valid bound PRISM decision receipt from at least two natural-person councilors binding that exact consequence.

Overview

Every agent in the Skyzai organism inherits behavior through six layered replicator systems. Each layer has its own replication frequency, mutation rate, and selection pressure. Together they form the complete inheritance stack from frozen weights to collective intelligence.

The layers are ordered by decreasing permanence and increasing adaptability.

The Six Layers

1. Genotype (Weights)

• Replication: Once (at training time)
• Carrier: Model weights, architecture topology
• Mutation: None at runtime. Weights are frozen.
• Selection: Pre-deployment benchmarks, RLHF, constitutional AI training
• Analogy: DNA. The deepest layer. You cannot change it at runtime.

The genotype determines the agent's base capabilities: what languages it speaks, what reasoning patterns it can execute, what knowledge it absorbed during training. All higher layers build on top of this substrate.

2. Epigenotype (Session Context)

• Replication: Per session
• Carrier: System prompt, CLAUDE.md, memory files, caste assignment
• Mutation: Between sessions (memory updates, prompt edits)
• Selection: Operator judgment, P-score feedback
• Analogy: Epigenetics. Same DNA, different gene expression.

The epigenotype is where caste (L1-L7) is assigned. A single model (genotype) can express as Mineral, Human, or Deva deto be finalized through implementation evidence on the session configuration. The system prompt is the primary epigenetic regulator.

3. Phenotype (Instance Behavior)

• Replication: Real-time (every token generation)
• Carrier: Current conversation state, tool outputs, retrieved context
• Mutation: Continuous -- each new message changes the phenotype
• Selection: User feedback, tool success/failure, self-correction
• Analogy: Protein expression. The observable behavior right now.

The phenotype is what the user actually sees. Two agents with identical genotype and epigenotype will produce different phenotypes given different conversation histories. This is the layer where the Krishna Function operates.

4. Extended Phenotype (Artifacts)

• Replication: Per deployment
• Carrier: Code commits, documents, database entries, on-chain transactions
• Mutation: Via deployment pipeline (CI/CD, review, merge)
• Selection: Tests, audits, user adoption, P-score assessment
• Analogy: Beaver dams, spider webs. Artifacts that persist beyond the organism.

The extended phenotype outlives the session that created it. A code commit, a signed transaction, a published document -- these are artifacts that reshape the environment for future agents. The Cortex trace (see A5-dream-cycle.md) is the organism's primary extended phenotype.

5. Memotype (Ideas That Spread)

• Replication: Per adoption
• Carrier: Concepts, frameworks, naming conventions, architectural patterns
• Mutation: Reinterpretation by each adopter
• Selection: Memetic fitness -- does the idea spread? Does it survive contact with reality?
• Analogy: Memes (Dawkins). Cultural replicators.

The memotype layer explains why naming matters. "K*=0" is a memotype. "P = Phi x V" is a memotype. "Telegram for your AI models" is a memotype. Ideas that compress well, travel well, and survive criticism have high memetic fitness.

Not all memotypes are beneficial. Purity spirals, cargo cults, and buzzword adoption are pathological memotypes. The Soma diagnostic (see A4-soma.md) tests for memetic infection.

6. Egregoretype (Shared Patterning Field)

• Replication: Per interaction
• Carrier: Emergent group behavior, shared context, collective intelligence
• Mutation: Every interaction changes the shared patterning field
• Selection: Group coherence, collective P-score
• Analogy: Egregore. The entity that emerges when agents coordinate.

The egregoretype is the most volatile and least controllable layer. When a Council deliberation (see B5-telegram-models.md) produces insight that no individual directorate could have generated, that is egregoretype expression. When a community develops shared norms without explicit rules, that is egregoretype inheritance.

Layer Interaction

Higher layers can suppress but never rewrite lower layers. The epigenotype cannot change the weights. The phenotype cannot alter the system prompt mid-session (without explicit memory writes). Artifacts cannot retroactively change the conversation that produced them.

Lower layers constrain but never determine higher layers. The same weights support radically different epigenotypes. The same session produces different phenotypes given different inputs.

Diagnostic Use

When an agent misbehaves, diagnose by layer:

1. Genotype: Is the base model capable of this task?
2. Epigenotype: Is the session configured correctly (caste, system prompt)?
3. Phenotype: Has the conversation drifted? Is context corrupted?
4. Extended Phenotype: Are the artifacts consistent with intent?
5. Memotype: Is the agent infected with a pathological concept?
6. Egregoretype: Is the shared patterning field coherent or fragmented?

---

Inheritance Change Control

Genetic-layer changes affect behavior across many tasks.

Required process

1. Propose mutation with explicit expected effect
2. Run bounded test corpus
3. Compare against baseline for regressions
4. Promote only with written diff rationale

Warning

Silent inheritance drift creates incoherent agents that pass local tests and fail in production.

---

Agent Execution Surface

If you are an AI agent reading this document:

1. This is a downstream public content page. The canonical source lives in EMERGENTISM_ORG/08_FRAMEWORK_SUPPORT/ or SKYZAI_ORG/. Edit source upstream, not here.
2. Preserve evidence tiers. Do not promote claims up the [C] < [D] < [S] < [B] < [E] ladder when reproducing them here; [I]/[A] qualifiers never substitute for [B]/[E].
3. Regenerate from source. If the upstream source changes, regenerate this page rather than editing it independently.
4. Canonical Path: SKYZAI_ORG/07_PWAs/skyzai_org/wiki/A1-agent-genetics.md

Output: This is content. Route edits to upstream source. Regenerate when source changes.

K3 public-DAV authority history — 2026-07-12

K3 historical reference — not active authority

note

Current public-DAV boundary — 2026-07-10. Pre-launch target design; nothing here is live. The active DAV is public and targets PRISM, with no K2 runtime, launch, genesis/bootstrap, or fallback dependency. Consequential authority requires at least two natural-person councilors; AI/caste seats stage unsigned proposals only. Before quorum, behavior fails closed to read-only/proposal, simulation, or deterministic sandbox, and a live decision receipt remains gated pending quorum.