Research Engine Guide

The Research Engine is an autonomous OSINT & MDP intelligence pipeline designed for multi-step web research, deep fact extraction, relational ecosystem discovery, and primary source validation. It formalizes web research as an active Markov Decision Process, guaranteeing high precision, anti-hedging discipline, and epistemic rigor.

Key Capabilities & Features

• Lossless Question Inventory: Deconstructs multi-part research prompts without dropping any specific metric, trade, or comparison point.
• 3-Part Query Triangulation Ensembles: Generates Canonical ID, Relational Ecosystem, and Friction/Signal queries for every concept, reducing retrieval variance by up to 22%.
• Active Anti-Hedging Retrieval: Bans vague generalities ("various partners", "industry standard") and extracts named entities down to exact numbers, registration IDs, and verified credits.
• Autonomous Gap-to-Query Loop: Translates missing data points into surgical Boolean repair queries, executing automated targeted secondary retrieval.
• 5-Tier Epistemic Fact Separation: Categorizes findings into Canonical Hard Facts, Industry Estimates, Relational Ecosystem, Community/Culture, and Self-Claims.
• Primary Source Verification Matrix: Concludes every research dossier with a 3-column verification matrix linking claims to verified primary URLs with zero broken citation tags.

How the Research Loop Works

The pipeline executes through five coordinated phases:

[ User Request / Research Target ] 
                  │
                  ▼
1. Mission Control (Lossless Inventory & 3-Part Query Ensembles)
                  │
                  ▼
2. Web Swarm (Active Multi-Spherical Live Search Retrieval)
                  │
                  ▼
3. Fact Compressor (Distillation, Relational Entities & Anti-Averaging)
                  │
                  ▼
4. Coverage Audit (Hard Depth Gatekeeper) ── (Incomplete?) ──► Back to Step 2
                  │ (Complete)
                  ▼
5. Master Synthesizer (360° Publication-Grade Intelligence Dossier)

1. Mission Control: Analyzes your research prompt, creates a numbered sub-question inventory, and formulates 3-part query triangulation ensembles covering all 6 Epistemic Spheres.
2. Web Swarm: Conducts live web searches across official registries, regulatory gazettes, specialized industry rankings, GitHub issues, and award databases.
3. Fact Compressor: Distills raw web data into clean, deduplicated empirical statements while preserving all relational entities and partner credits.
4. Coverage Audit: Evaluates whether all 6 spheres and user sub-questions contain sufficient empirical backing. If data gaps remain, it translates them into specific Boolean queries and triggers a targeted second search pass.
5. Master Synthesizer: Compiles the complete research dossier with BLUF executive summaries, 3-tier topic analysis, relational partner matrices, responsible domain heuristics, and a final verification matrix.

The 6 Universal Epistemic Spheres

1. Sphere 1 [Primary Ground Truth]: Canonical registers, official gazettes, court filings, peer-reviewed papers, RFC specs, test suites.
2. Sphere 2 [Specialized Trade & Industry Media]: Domain journals, benchmark databases, industry rankings, tech blogs, specialized press.
3. Sphere 3 [Audits, Certifications & Awards]: State awards, ISO/CVE registries, compliance certificates, institutional grants, peer audits.
4. Sphere 4 [Relational Ecosystem & Dependencies]: Subcontractor studios, production partners, joint ventures, co-authors, dependencies.
5. Sphere 5 [Grassroots, Culture & Practitioner Signals]: Developer issue trackers, employee sentiment (kununu, Glassdoor), post-mortems, community sentiment.
6. Sphere 6 [People, Leadership & Strategic Footprint]: Founders, PIs, lead architects, keynote transcripts, interview statements.

How to Run Research Engine

• Navigate to /drop/research or /drop/deepresearch
• Or select Research Engine from the Recipes catalog in the Reactor workspace.