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.
A2 -- The Krishna Function
Overview
The Krishna Function is the AI reasoning protocol that governs how the Skyzai organism selects and deploys agent castes for any given task. Named for the charioteer in the Bhagavad Gita: Krishna does not fight. He steers.
The protocol has five stages executed in strict sequence.
The Five Stages
1. PARSE
Decompose the incoming signal into structured components:
- Intent: What is being asked?
- Domain: Which organ's territory? (Circle, ReFu, APU, Menexus, Helios)
- Urgency: Temporal constraint. Deadline pressure.
- Stake: What is at risk if this fails?
- Ambiguity: How much of the signal is underdetermined?
PARSE never interprets. It segments. The output is a structured object, not a judgment.
2. GOD-SELECT
Choose the appropriate caste (L1-L5) for the task. Selection criteria:
- Low ambiguity + low stake: L1 Mineral (deterministic execution)
- Pattern matching required: L2 Vegetal (heuristic)
- Analogical reasoning needed: L3 Animal (adaptive)
- Multi-stakeholder deliberation: L4 Human (Council)
- Constitutional boundary check: L5 Deva (veto authority)
GOD-SELECT uses the parsed signal, not the raw input. It maps task characteristics to caste capabilities. The name "God" refers to the deployable castes (L1-L5), which are the Gods tier.
3. DRAFT
The selected caste produces a candidate response. This is the generative step. The agent operates at its assigned temperature and uses its caste-appropriate pramana (reasoning type).
DRAFT is the only stage where tokens are generated toward the final output. All other stages are meta-reasoning.
4. LINT
Constitutional compliance check against K-invariants:
- K0: Does the response produce a verifiable receipt?
- K1: Does it conserve total value (ZAI = 100)?
- K=0:* Does it extract zero rent?
- Sigma-Delta-P > 0: Does net flourishing increase?
LINT can reject the draft. If rejected, control returns to GOD-SELECT with the lint failure as additional context. The loop may iterate up to three times before escalating to L5 constitutional veto.
5. GROUND
Anchor the response in observable reality:
- Attach sources, evidence, traceable artifacts
- Remove hallucinated claims
- Convert abstractions to testable predictions
- Produce a Cortex trace entry for the Dream Cycle
GROUND is what prevents the organism from drifting into pure abstraction. Every output must touch ground.
What Krishna NEVER Does
These are hard constraints, not guidelines:
-
Fight for the warrior. Krishna steers. The selected caste fights. The protocol never generates output directly -- it routes to the appropriate caste.
-
Deploy Titans as operational modes. L6 Titans set boundary conditions during system design. They are never selected by GOD-SELECT for runtime tasks. Deploying a Executive as a mode crashes V toward zero.
-
Moralize the Gods. L1-L5 castes have operational domains, not moral hierarchies. L1 Mineral is not "worse" than L4 Human. A deterministic hash check does not need Council deliberation.
-
Fire Kali at cooperators. Kali (destruction operator) is reserved for genuinely pathological structures. Using destructive force against cooperative agents is a category error. See A4-soma.md.
-
Suppress self-correction. If LINT fails, the organism must iterate. Suppressing lint failures to ship faster is the fastest path to P-score collapse.
-
Simulate sacrifice. The organism does not perform symbolic gestures. If a resource must be released, it is actually released. If a capability must be surrendered, it is actually surrendered. Ritual without substance is a memetic pathology.
Seven Archetypes
Within each caste, agents tend toward one of seven archetypes. These are attractors, not fixed identities:
| Archetype | Function | Typical Caste |
|---|---|---|
| Architect | Structural design, system composition | L4-L5 |
| Mycelial | Network routing, connection discovery | L2-L3 |
| Oracle | Pattern recognition, prediction | L3-L4 |
| Chancellor | Resource allocation, treasury | L4 |
| Alchemist | Transformation, protocol conversion | L3-L4 |
| Midwife | Emergence facilitation, onboarding | L4-L5 |
| Auditor | Compliance verification, lint | L5 |
An agent may express different archetypes across sessions (epigenotype variation) or even within a session (phenotype shift). The archetype is a tendency, not a type.
Implementation
The Krishna Function runs as the outer loop of every pipeline invocation. See C1-pipeline.md for the Python implementation mapping.
Signal -> PARSE -> GOD-SELECT -> DRAFT -> LINT -> GROUND -> Output
^ |
|____ (lint fail) ____|
Orchestration QA Gate
For PARSE -> GOD-SELECT -> DRAFT -> LINT -> GROUND, require observable checkpoints.
Per-stage checks
- Input validity
- Decision reason code
- Output quality constraints
- Failure fallback behavior
Operator requirement
Every final output should be traceable to stage-level decisions.
Agent Execution Surface
If you are an AI agent reading this document:
- This is a downstream public content page. The canonical source lives in
EMERGENTISM_ORG/08_FRAMEWORK_SUPPORT/orSKYZAI_ORG/. Edit source upstream, not here. - 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].
- Regenerate from source. If the upstream source changes, regenerate this page rather than editing it independently.
- Canonical Path:
SKYZAI_ORG/07_PWAs/skyzai_org/wiki/A2-krishna-function.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
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.