بسم الله الرحمن الرحيم
Why IPSC
The only computationally graded hadith corpus in existence.
Every row shows the same capability. Left: what the field offers today. Right: what the IPSC provides.
Up to ~300K hadith across the largest platforms. Text digitized from printed editions.
449,285 hadith from 86 classical works — computationally parsed, not just digitized.
Grades copied from printed books or assigned by a single editorial team. No methodology published.
Every grade computed algorithmically via the five-condition Ibn al-Salah framework. Full methodology published.
Chains displayed as plain text. No structural parsing. No narrator identification.
Every chain parsed into individual narrator positions. 1.82 million positions extracted and resolved.
No canonical narrator identification. The same narrator in different collections is unlinked.
27,118 unique Person IDs. The same narrator across 5,000 chains always resolves to one identity.
No computational matn criticism. Quran morphology analysis exists (QAC) but not applied to hadith.
Two-pass matn criticism: Quran contradictions, fabrication patterns, anachronistic vocabulary, prophetic baseline.
Collections exist in silos. No linking of the same hadith across different books.
52,938 matn clusters linking parallel narrations across all 86 collections with attestation levels.
No teacher-student graph. No network analysis capability.
889,913 directed edges mapping 1,400 years of who-learned-from-whom across 7,973 narrator nodes.
Grades presented without reasoning. Some cite a scholar name without tracing the chain of evidence.
Every grade traces to a specific NRS tier, a Taqrib al-Tahdhib entry number, and a documented grading rule.
No published testing process. No regression suite. No adversarial validation.
Continuous quality cycle: 33-test regression suite + 7-test cross-field consistency check + NAQD-1/2/3 adversarial framework, all gating release. v3.9 LLM-tiebreaker decisions cite the structural validator chain (tahdīb graph + tabaqah + temporal), with provenance stamped on every decision.
No IP claims. No filed protections.
US patent applications filed for the matn-criticism pipeline, NRS synthesis, classical-dimension detectors, and semantic cluster + chain-matn conflict detection. IPSC™ / MindHYVE™ / Eve-Theology™ trademarks.
No external review. No published audit practice.
Independent external examiners pressure-test framing on a recurring cadence; the audit-practice page at /honesty documents pillars, recalibration timeline, and the open scholar-collaboration program.
Keyword search only.
11 vector indexes, 3,072-dim OpenAI embeddings. Semantic similarity over matn (3 tiers), narrators (3 tiers), defects (2 tiers), and glossary terms (3 tiers). “Find me terms about chain weakness”.
Each of these required original engineering grounded in classical hadith methodology.
Mudallis Detection
105 narrators catalogued with severity 1–5. When level 3+ uses ambiguous transmission, the chain is flagged. 14,379 positions carry requiresTasrih.
Ikhtilat Registry
67 narrators with documented mental deterioration. Onset years, pre/post student lists. 116,528 hadith carry structured ikhtilat data.
Fabrication Validation
6,822 known fabrications graded through the same engine. The system correctly identifies what scholars identified centuries ago.
Quran Cross-Reference
1,279,676 term matches and 6,447 direct quotation matches between hadith and Quranic verses.
Taqwiyah with Integrity
Supporting chains verified for independence. T10+ narrators can never be elevated. 168,284 upgrades applied with safeguards.
Provenance Trail
Every correction cites a Taqrib entry, Tahdhib reference, or named scholarly source where possible. Honest exception: v3.9 multi-stage validator + v3.26 LLM tiebreaker decisions cite the structural validator chain (tahdīb graph + tabaqah + temporal), not a named imām naṣṣ — those records carry method: 'v3.26-llm-tiebreaker-sonnet' provenance. See the disclosure.
Entity Layer
54,885 entities aggregate all chains, collections, and grades per teaching. One teaching, all its evidence, one record.
Hadith Hawala Splitting
8,258 compound chains identified by the ح marker. Each branch graded independently.
ما يمكن للباحثين فعله الآن
Queries that took months — now take seconds
"Find every hadith narrated through Hammad ibn Salamah"
One PID query returns every chain across all 86 collections.
"Flag every chain where Qatadah uses 'an'anah"
Mudallis detection on 14,379 positions — instantly.
"Reclassify a narrator — how many grades change?"
Instant impact analysis via the provenance trail.