Skip to content

Q001-H3 — Partial Implementations Exist

Research R0049 — Landscape Scan
Run 2026-03-31-02
Query Q001
Hypothesis H3

Statement

Partial implementations exist that apply individual structured analytic techniques or research workflow stages via LLMs, but no published prompt achieves comprehensive coverage of a full analytical rigor framework. The gap between narrow-task prompts and full-framework prompts remains unfilled.

Status

Supported — This is the best-supported hypothesis based on the evidence.

Supporting Evidence

Evidence ID Summary Strength
SRC02-E01 sroberts implements 3 of 66 SATs (Starbursting, ACH, Key Assumptions Check) Strong
SRC05-E01 Agent Laboratory: multi-phase research pipeline in code, not prompt Strong
SRC06-E01 AI-Researcher: complete research workflow, NeurIPS 2025 Strong
SRC01-E01 58 prompting techniques catalogued, none a full research framework Strong
SRC03-E01 PRISMA-trAIce: reporting standard for AI in reviews, not an operational prompt Moderate
SRC04-E01 Commercial prompts prioritize citation transparency over analytical rigor Moderate

Contradicting Evidence

Evidence ID Summary Strength
No direct contradicting evidence

Reasoning

The evidence converges on a clear pattern: the field has produced (a) numerous narrow-task prompts for specific review stages (screening, extraction, checklist verification), (b) a handful of partial implementations of structured analytic techniques via LLMs, and (c) comprehensive research agent systems implemented in code rather than as system prompts. The gap is specifically at the intersection of "published as a system prompt" and "implements a full analytical rigor framework." This gap likely exists because comprehensive methodologies require multi-step orchestration that exceeds what a single system prompt can effectively encode — hence the shift to code-based agent architectures.

Relationship to Other Hypotheses

  • Refines both H1 (too optimistic) and H2 (too absolute)
  • Best explains the observed evidence landscape