Skip to content

R0049/2026-03-31

Research R0049 — Landscape Scan
Mode Query
Run date 2026-03-31
Queries 3
Prompt Unified Research Standard v1 — Query Mode
Model Claude Opus 4.6 (1M context)
Q001 — AI Research Prompt Frameworks

Has anyone published a complete, usable AI/LLM system prompt that implements a full analytical rigor framework for research?

Very unlikely (05-20%) that a complete published framework exists. Confidence: Medium-High.

Supported: H3 — Partial implementations exist but none achieve comprehensive coverage

H1 — Full framework exists — Partially supported

H2 — Nothing exists — Eliminated

Full query results

Q002 — Unified IC-Scientific Methodology

Has anyone published a systematic combination of intelligence community analytical standards with scientific methodology frameworks?

Very unlikely (05-20%) that such a combination exists. Confidence: Medium.

Supported: H2 — No combination published

Supported: H3 — Comparison without integration

H1 — Unified methodology exists — Eliminated

Full query results

Q003 — AI Research Tools with Structured Frameworks

What AI-assisted research tools implement structured analytical frameworks beyond simple chat-based research?

Very unlikely (05-20%) that a comprehensive structured tool exists undiscovered. Confidence: Medium-High.

Supported: H3 — Isolated features only, no comprehensive tool

H1 — Multiple comprehensive tools exist — Eliminated

H2 — No features at all — Eliminated

Full query results

Collection Analysis

Cross-Cutting Patterns

The three queries converge on a single finding: the integration of formal analytical rigor methodology into AI research systems is a gap across every layer of the stack — from theoretical methodology (Q002) to operational prompts (Q001) to production tools (Q003).

Specific patterns observed:

  1. Efficiency over rigor: The AI research ecosystem universally optimizes for speed, volume, and citation accuracy. Analytical rigor features (competing hypotheses, calibrated probability, self-audit) are absent from all major tools.

  2. Parallel worlds: Intelligence community analytical standards (ICD 203, structured analytic techniques) and scientific methodology frameworks (GRADE, PRISMA, Cochrane, IPCC) have developed remarkably similar structures independently but have never been combined.

  3. Partial implementations are siloed: The few partial implementations found (Roberts' LLM SATs, Framework CoT, scite Smart Citations, Open Synthesis ACH, MS Copilot Critique) each address a single technique or feature in isolation with no integration between them.

  4. The methodology gap drives the tools gap: Since no unified IC-scientific methodology exists (Q002), tools have no framework to implement (Q003), and prompts have no methodology to encode (Q001).

Collection Statistics

Metric Value
Queries 3
Hypotheses generated 9 (3 per query)
Hypotheses supported 4
Hypotheses partially supported 1
Hypotheses eliminated 4
Searches executed 9
Total results returned ~150
Results selected 22
Results rejected ~128
Sources evaluated 16
Evidence items 16

Source Independence

Sources across the three queries are highly independent:

  • Q001 sources: Academic publications (PRISMA-trAIce, Framework CoT), practitioner blog (Roberts), GitHub repositories (LLM Prompt Library, Agent Laboratory)
  • Q002 sources: Peer-reviewed journals (Duke, IPCC, PLOS ONE), institutional report (RAND)
  • Q003 sources: Open-source repositories (PaperQA2, STORM, GPT Researcher, Open Synthesis), commercial products (Elicit, scite), news coverage (MS Copilot Critique)

No source appears in more than one query. The convergent finding (gap across all layers) emerges from independent evidence streams.

Collection Gaps

Gap Queries affected Impact
Classified/proprietary IC research Q001, Q002 Could contain unpublished framework integration
Enterprise tools (Palantir, Maltego) Q003 Proprietary features not verifiable
Non-English publications Q001, Q002 Minor — field is English-dominated
Dissertations/theses Q002 Most likely venue for novel cross-domain proposals
AI company internal research Q001, Q003 Anthropic, OpenAI, Google may have internal methodology prompts

Collection Self-Audit

The consistent finding across all three queries — that comprehensive analytical rigor frameworks are absent from the AI research landscape — must be evaluated against the researcher's incentive to find this result (as the builder of such a framework). Three mitigation measures were applied:

  1. Searches designed to find confirming evidence: All search strategies were optimized to discover existing frameworks, not to confirm their absence.
  2. Generous evaluation of partial implementations: Roberts' SATs, Framework CoT, scite Smart Citations, and MS Copilot Critique were all documented as genuine prior art rather than dismissed.
  3. Transparent documentation of limitations: All search scope limitations (classified research, proprietary tools, non-English publications) are documented with their potential impact on confidence.

Overall collection risk: Low with some concerns — primarily about inaccessible proprietary and classified sources.

Resources

Summary

Resource Value
Web searches 18
Web fetches 5
Files produced 90
Duration (wall clock) 30m 52s
Tool uses (total) 137

Tool Breakdown

Tool Count
WebSearch 18
WebFetch 5 (GitHub repos: PaperQA2, STORM, Open Synthesis, GPT Researcher, LLM Prompt Library)

Token Distribution

Phase Approximate share
Search execution 30%
Source evaluation 20%
File generation 50%