PUMA Vault

🧠 MOC — Methods & Frameworks

Se lee en 5 min

🧠 MOC — Methods & Frameworks

Overview

All research, development, and knowledge management methodologies integrated in PUMA and this vault.

📊 Research Methodologies

Design Science Research (DSR)

  • PN-DSR-SLR-Methods — DSR paradigm explanation
  • Core idea: research must produce and evaluate useful artefacts. PUMA artefact = benchmark framework.
  • SRC-MITAILab-WP316 — MIT AI Lab WP316 (active research methods)
  • Reference: Hevner et al. (2004), Peffers et al. (2007)

Systematic Literature Review (SLR + PRISMA)

Statistical Methods

Reference Class Forecasting (RCF)

🤖 Development Methodologies

Spec-Driven Development (SDD)

OpenSpec

  • PN-SDD-Framework — OpenSpec section
  • Best for: brownfield / iterative spec updates

Spec Kit (GitHub)

  • PN-SDD-Framework — Spec Kit section
  • Best for: greenfield with strict phase gates + constitution.md

BMAD (Multi-Agent Simulation)

Context-Driven Development (CDD)

📚 Knowledge Management Methodologies

PARA (Projects, Areas, Resources, Archive)

GTD (Getting Things Done)

Zettelkasten

Johnny Decimal

  • JD-Master-Index — Full index
  • Implemented in: all folder numbers (10–90, decimal sub-IDs)

Maps of Content (MOCs)

🤖 AI Prompting Methodologies

Keshav Three-Pass Method

RCOIF Framework

  • PN-RCOIF-Framework — Full explanation
  • Role · Context · Objective · Instructions · Format

Chain-of-Thought (CoT) + Zero-Shot CoT

One-Shot and Few-Shot Prompting

CO-STAR + Self-Consistency + Structured Output

Contextual Anchoring

EGI (Exploración Guiada Interactiva)

  • PN-EGI-Framework — Full explanation
  • Used for: domain exploration in new literature areas

AMI (Autodiagnóstico y Mejora Iterativa)

DRCA (Deconstrucción y Reconstrucción Conceptual Avanzada)

IIPR (Ingeniería Inversa de Prompts y Respuestas)

Reflexion (Verbal Self-Reflection Loop)

Tree of Thoughts (ToT)

  • PN-TreeOfThoughts-Deliberate — BFS/DFS search over reasoning steps; PUMA Stage 3 sprint planning
  • LN-Yao-2023-TreeOfThoughts — Source: Yao et al. (NeurIPS 2023)
  • Key distinction from CoT: supports backtracking; 18.5× improvement on multi-step problems; 50–200× compute cost

Context Engineering (CE)

  • PN-ContextEngineering — Six-slot context pipeline; token budget allocation; context pollution mitigation
  • LN-Videos-Context-Engineering — Video references
  • CE = 2026 successor to Prompt Engineering: design the information system that fills the context, not just the prompt text

RLHF and Constitutional AI

  • PN-RLHF-Constitutional — RLHF pipeline, DPO, Constitutional AI, RLAIF; alignment method per PUMA model
  • Alignment training paradigm for all PUMA models (Llama-Instruct, Claude Sonnet, DeepSeek-R1)

Model Context Protocol (MCP)

Transformer Architecture and MoE

  • PN-Transformer-MoE — Self-attention, GQA, RoPE, SwiGLU, MoE routing, KV cache; PUMA model technical reference
  • LN-Fedus-2022-SwitchTransformers — MoE source: Fedus et al. (JMLR 2022)
  • Foundation for understanding DeepSeek-V3 MoE efficiency rationale (671B total / 37B active)

Marco Veritas (AI Transparency & Academic Integrity)

  • PN-Veritas-Framework — Full framework: proactive disclosure, primary source verification, no delegation of judgement
  • AI-Use-Log — PUMA AI use log (PRISMA-trAIce compliance)
  • Author: Codina (2024) · Applied in: PUMA Constitution §7, BMAD QA Agent audit

AI Code Quality (Empirical Evidence)

  • LN-CodeRabbit-2025-AIvsHumanCode — CodeRabbit (2025): 470 PRs, AI 1.7× more issues, 2.74× security, 3× readability; 7 mitigation strategies mapped to PUMA design

🏢 Business Systems & Operations Management

Systems Thinking & Process Documentation

Business Operating Systems

  • LN-Wickman-2012-Traction — Traction / EOS (Wickman, 2012): 6-component operating system (Vision, People, Data, Issues, Process, Traction)
  • LN-Harnish-2022-ScalingUp — Scaling Up (Harnish, 2022): Rockefeller Habits, Four Decisions, One-Page Strategic Plan

Theory of Constraints (TOC)

  • LN-Goldratt-2004-TheGoal — The Goal (Goldratt & Cox, 2004): TOC, Five Focusing Steps, Drum-Buffer-Rope, throughput accounting
  • Core concept: identify bottleneck → exploit → subordinate → elevate → repeat

Customer Experience & Frictionless Design

  • LN-Price-2022-Frictionless — The Frictionless Organization (Price & Jaffe, 2022): Customer Effort Score, friction mapping, DIRTFT, five demand types

🔗 Orphan Check

LIST
FROM "30 - Permanent/33 Frameworks"
WHERE file.name != this.file.name
SORT file.name ASC

Vista Gráfica

Retroenlaces