Definition

NABLA Infinity (written as the mathematical symbol for the gradient operator, suggesting infinite-dimensional exploration of epistemic space) is the foundational epistemic framework governing all belief formation and knowledge validation within the Prismatic Platform. It defines seven non-negotiable axioms that every agent, pipeline, and decision process must satisfy. Five axioms carry hard enforcement (violations trigger immediate blocking), and two carry soft enforcement (violations trigger warnings and investigation).

The framework operates at 3NL Level 3 integration, meaning it is not a suggestion or a best practice but a structural constraint woven into the platform's runtime. Agents cannot form beliefs, validate claims, or make decisions without NABLA axiom compliance. The axioms are enforced programmatically through the epistemic pipeline, validated through QEVE, and gated through Trinity Gate.

The core philosophical position of NABLA Infinity is uncompromising: Reality is not a democracy. Evidence is not optional. Contradictions are not embarrassments. The framework treats epistemic hygiene with the same rigor that type systems bring to software correctness -- violations are caught at the structural level, not left to human judgment.

The 7 Axioms

Axiom 1: Signal Plurality (HARD Enforcement)

Requirement: A minimum of two independent signals is required before any belief can be formed.

No claim, hypothesis, or conclusion may rest on a single source of evidence, regardless of how authoritative that source appears. Single-source beliefs are structurally fragile -- if the source is wrong, compromised, or stale, the entire belief collapses with no warning signal.

Signal Plurality enforcement validates independence at the source level. Two articles from the same news outlet referencing the same press release count as one signal, not two. Two independent investigations reaching the same conclusion from different methodologies count as two signals. The independence_group field in the evidence data model tracks which signals share common origins.

Violation response: E2 BLOCK. The belief is rejected until additional independent signals are provided. No bypass exists for this axiom.

Axiom 2: Contradiction Preservation (HARD Enforcement)

Requirement: When contradictory evidence exists, both sides must be preserved explicitly. Neither side may be discarded, downweighted, or rationalized away.

This is the most counterintuitive axiom and the one most frequently violated by conventional AI systems. When Signal A says "risk is high" and Signal B says "risk is low," the natural impulse is to resolve the contradiction -- pick a winner, compute an average, or dismiss the weaker signal. NABLA Infinity forbids all of these responses.

Contradictions are preserved because they carry information. A contradiction between two credible sources indicates either (a) the situation is genuinely ambiguous, (b) one source has information the other lacks, or (c) the framing of the question is flawed. All three possibilities are valuable. Premature resolution destroys this information.

Preserved contradictions are annotated with severity (weak, moderate, strong) and propagated through the pipeline. The QEVE confidence formula includes a contradiction index that reduces final confidence in proportion to unresolved contradictions, ensuring they are never ignored even if they are preserved.

Violation response: E2 BLOCK. Any operation that discards contradictory evidence is halted immediately.

Axiom 3: Absence Informative (SOFT Enforcement)

Requirement: Missing signals must be tracked as data points. The absence of expected evidence is itself evidence.

When a company that should have public financial filings has none, the absence is informative. When a domain that should have SSL certificates lacks them, the absence is a signal. NABLA Infinity requires that expected-but-missing evidence be recorded with the same rigor as present evidence, tracked through the belief graph, and factored into confidence calculations.

Absence tracking prevents a common failure mode: concluding that "no evidence of risk" means "evidence of no risk." These are logically distinct statements, and NABLA Infinity enforces the distinction.

Violation response: E1 WARNING. Investigation triggered. Absence tracking failures are less immediately dangerous than active evidence suppression but can accumulate into significant blind spots.

Axiom 4: Time Decay (HARD Enforcement)

Requirement: All beliefs must carry mandatory timestamps, and evidence weight must decay over time according to configurable decay functions.

Evidence gathered six months ago is less reliable than evidence gathered today. Regulatory environments change. Corporate structures are restructured. Threat landscapes evolve. NABLA Infinity requires that every signal carry a collection timestamp and that belief strength decreases automatically as evidence ages.

Decay functions are configurable per domain. Financial data may decay faster than geological data. Threat intelligence decays faster than corporate registration records. The decay rate is a domain-specific parameter, but the requirement for decay is universal.

Time decay prevents a specific failure mode: building high-confidence conclusions on stale evidence. A due diligence assessment that relies on 18-month-old data without accounting for staleness is operating on a false sense of precision.

Violation response: E2 BLOCK. Beliefs without timestamps are rejected. Beliefs with timestamps but no decay applied are rejected.

Axiom 5: Unknown Valid (HARD Enforcement)

Requirement: "I don't know" is a legitimate epistemic state. The system must be capable of expressing and propagating uncertainty rather than forcing a determination.

This axiom directly counters the tendency of AI systems to always produce an answer. When evidence is insufficient, contradictory, or stale, the correct response is not a low-confidence guess but an explicit statement of uncertainty. NABLA Infinity requires that the platform's agents, pipelines, and outputs support an "unknown" state with the same structural rigor as "true" and "false."

Unknown states propagate through the belief graph. If a critical premise is unknown, conclusions that depend on it inherit the unknown state rather than silently defaulting to an assumed value. This propagation ensures that uncertainty is visible to decision-makers rather than buried under layers of inference.

Violation response: E2 BLOCK. Any operation that forces a determination when evidence is insufficient is halted.

Axiom 6: Source Independence (SOFT Enforcement)

Requirement: Independent sources must be weighted higher than correlated sources. Source correlation must be tracked and factored into confidence calculations.

Ten articles from different outlets citing the same original report represent one source, not ten. NABLA Infinity requires that source independence be explicitly tracked through independence grouping and that correlated sources be appropriately discounted in confidence calculations.

Source independence is particularly critical in OSINT (Open Source Intelligence) contexts where information amplification is common. A single government press release can generate dozens of news articles, social media posts, and analyst reports, all of which trace back to the same original source. Without independence tracking, the illusion of plurality masks a single-source belief.

Violation response: E1 WARNING. Bias assessment required. Source correlation that is tracked but not corrected receives a warning rather than a block, as the tracking itself provides transparency.

Axiom 7: Provenance Mandatory (HARD Enforcement)

Requirement: All beliefs must be traceable to their origin signals through a complete, auditable chain of custody.

Every conclusion in the platform must answer the question: "Where did this come from?" The provenance chain links each belief to the evidence that supports it, the inference rules that derived it, and the raw data from which the evidence was extracted. Provenance enables three critical capabilities:

Audit: External reviewers can trace any conclusion back to its source data
Debugging: When a conclusion is wrong, provenance identifies which evidence or inference step failed
Accountability: Provenance establishes who (or what agent) produced each component of the reasoning chain

Provenance is enforced at the data model level. Every evidence struct includes a provenance field (full chain of custody) and a raw_data_hash (cryptographic hash of the original data for integrity verification). Evidence without provenance is rejected before entering the pipeline.

Violation response: E2 BLOCK. Beliefs without traceable provenance are rejected. No bypass exists.

Axiom Summary

#	Axiom	Enforcement	Violation	Core Principle
1	Signal Plurality	HARD	E2 BLOCK	No single-source beliefs
2	Contradiction Preservation	HARD	E2 BLOCK	Never discard inconvenient evidence
3	Absence Informative	SOFT	E1 WARNING	Missing evidence is evidence
4	Time Decay	HARD	E2 BLOCK	Beliefs weaken over time
5	Unknown Valid	HARD	E2 BLOCK	"I don't know" is acceptable
6	Source Independence	SOFT	E1 WARNING	Correlated sources are not independent
7	Provenance Mandatory	HARD	E2 BLOCK	All beliefs must be traceable

Addiction Preservation

Addiction Preservation is the platform's commitment to preserving contradictory signals, maintaining evidence plurality, and refusing to "smooth over" inconvenient truths. The term is deliberate: like addiction recovery, epistemic hygiene requires constant vigilance against the human tendency to rationalize, dismiss, or cherry-pick evidence.

The analogy is precise. An addict in recovery must remain vigilant against triggers that seem harmless but lead to relapse. Similarly, an epistemic system must remain vigilant against seemingly reasonable shortcuts that erode rigor:

"This contradiction is probably just noise -- let's remove it." (Contradiction Burial)
"We only need one more source to confirm." (Confirmation Bias)
"The timestamp is close enough -- no need for decay." (Time Decay bypass)
"Everyone knows this is true." (Single Source masked as common knowledge)

Addiction Preservation mandates that the platform treat each of these impulses as a threat to epistemic integrity, not as a reasonable optimization. The cost of preserving a spurious contradiction is low (one extra node in the belief graph). The cost of discarding a genuine contradiction is potentially catastrophic (a false conclusion treated as verified truth).

Anti-Patterns

NABLA Infinity defines five forbidden anti-patterns. These are not guidelines but hard enforcement rules that trigger blocking or halting when detected:

Anti-Pattern	Description	Axioms Violated	Enforcement
Cherry Picking	Selecting only evidence that supports a predetermined conclusion while ignoring or downweighting contradictory evidence	Signal Plurality, Contradiction Preservation	E2 BLOCK
False Certainty	Presenting a conclusion with higher confidence than the evidence supports, typically by ignoring uncertainty or absence signals	Unknown Valid, Time Decay	E2 BLOCK
Contradiction Burial	Acknowledging a contradiction exists but structurally hiding it so downstream consumers never see it	Contradiction Preservation	E3 HALT
Single Source Truth	Building a belief on one source (or multiple correlated sources disguised as independent)	Signal Plurality, Source Independence	E2 BLOCK
Reasoning Opacity	Producing conclusions without a traceable reasoning chain, making audit impossible	Provenance Mandatory	E2 BLOCK

Contradiction Burial receives the most severe enforcement (E3 HALT rather than E2 BLOCK) because it is the most dangerous anti-pattern. Unlike Cherry Picking, which is detectable through signal analysis, Contradiction Burial actively hides evidence within the system's own data structures. An E3 HALT requires mandatory review at supreme authority level before operations can resume.

Enforcement Protocol

NABLA Infinity enforcement operates at four escalating levels:

Level	Trigger	Response	Authority Required
E1	Single soft axiom violation (Absence Informative or Source Independence)	Warning issued. Correction requested. Operation continues with logged deviation	Agent-level
E2	Single hard axiom violation or anti-pattern detection	BLOCK. Operation halted. Rejection issued. Must be corrected before proceeding	System-level
E3	Trinity Gate failure or Contradiction Burial detection	HALT. Mandatory review required. No bypass. Full diagnostic logged	Supreme authority
E4	Multiple simultaneous axiom violations or systematic pattern of violations	Full investigation. Audit of all recent evaluations. Root cause analysis required	Cosmic clearance

Enforcement is non-bypassable. There is no flag, configuration option, or authority level that can disable axiom enforcement. The axioms are structural constraints, not policy preferences. Disabling them would be equivalent to disabling type checking in a compiled language -- technically possible but architecturally destructive.

Integration with Trinity Gate

NABLA Infinity and Trinity Gate serve complementary but distinct roles in the epistemic pipeline:

NABLA Infinity governs belief formation: How evidence is gathered, structured, weighted, and connected into a belief graph
Trinity Gate governs belief acceptance: Whether a formed belief meets the structural, logical, and formal standards required for acceptance

The relationship is sequential. NABLA axioms constrain the input to Trinity Gate. A belief graph that violates Signal Plurality will fail the Trinity Gate's logical consistency check. A belief graph that buries contradictions will fail the structural consistency check. NABLA compliance is a necessary (but not sufficient) condition for Trinity Gate passage.

QEVE sits between the two, providing the verification machinery that translates NABLA-compliant belief graphs into Trinity Gate evaluations. The QEVE pipeline's first stage (Graph Build) directly enforces NABLA axioms, while its later stages feed into Trinity Gate's three checks.

Transition to Execution

The transition from NABLA-governed exploration to NM/ND-governed execution follows a strict protocol:

EXPLORATION PHASE (NABLA Infinity active)
  - Maps uncertainty across the evidence space
  - Preserves all contradictions without resolution
  - Maintains parallel hypotheses simultaneously
  - Computes confidence scores with full uncertainty quantification
  - Tracks time decay and provenance for all signals
        |
        v
TRANSITION CONDITIONS (ALL must be satisfied)
  - confidence >= 0.95 (for critical decisions)
  - trinity_gate.passed (all 3 layers + meta-integrity)
  - axioms_compliant (all 7 NABLA axioms satisfied)
        |
        v
EXECUTION PHASE (NM/ND Doctrine active)
  - [No Mercy](@/glossary/no-mercy.md): Complete execution, zero tolerance for incomplete delivery
  - [No Doubts](@/glossary/no-doubts.md): Full commitment to the verified conclusion
  - Decisive action based on formally verified, robustness-tested beliefs

The transition is the critical moment in the platform's reasoning process. Before the transition, doubt is not just acceptable but mandatory -- NABLA requires preserving uncertainty, maintaining contradictions, and acknowledging the unknown. After the transition, doubt is replaced by commitment -- the NM/ND doctrine requires complete execution without hedging.

This dual-phase approach prevents both failure modes: acting too soon on unverified beliefs (premature execution) and never acting because there is always more evidence to gather (analysis paralysis). NABLA's axioms define what "sufficient verification" means. Trinity Gate certifies that verification is complete. NM/ND governs what happens next.

Trinity Gate -- 4-layer verification gate for NABLA-governed beliefs
QEVE -- Verification engine implementing NABLA axioms in its pipeline
Signal Plurality -- Core axiom requiring minimum 2 independent signals
Confidence Threshold -- Decision thresholds derived from NABLA axiom compliance
Provenance Mandatory -- Axiom requiring full traceability for all beliefs
NM/ND Doctrine -- Execution doctrine activated when NABLA confidence is sufficient
No Mercy -- Execution-phase enforcement of complete delivery
No Doubts -- Execution-phase enforcement of decisive action
3NL Framework -- Integration framework connecting NABLA to AIAD agents
Epistemic Pipeline -- The 16-level pipeline in which NABLA operates
White Team -- Verification team producing NABLA-compliant formal proofs
Red Team -- Adversarial team testing NABLA's resistance to epistemic attacks
Blue Team -- Defensive team maintaining NABLA compliance across the platform
Gray Team -- Boundary exploration team surfacing NABLA edge cases
Lean4 -- Theorem prover used in formal verification of NABLA-governed beliefs

Connect & Contribute

Created by Tomáš Korcak (korczis) | Open Source under GHL

GitHub | GitLab | LinkedIn | Contact
Developer Portal | Architecture | Meet the Creator

NABLA Infinity