research-first-developmentlisted

# Research-First Development > Build knowledge bases that build software — research before code, teach before execute Methodology for AI-assisted development where investment in research, skill creation, and knowledge encoding **precedes** implementation. Discovered through the Dead Letter heir's masterclass on cognitive architecture utilization (February 2026). --- ## When to Use - Starting any new project with the AI assistant - Entering a new implementation phase of an existing project - Onboarding the AI assistant to an unfamiliar domain - When AI output quality is inconsistent (symptom: insufficient context) - Before any complex multi-subsystem implementation --- ## Core Insight Traditional software development: **Requirements → Design → Code → Test** AI-assisted development with cognitive architecture: **Research → Teach → Plan → Execute** > **The quality of AI output is directly proportional to the quality of knowledge in its context.** Instead of the human writing code with AI assistance, **the human orchestrates intent while AI handles execution**. But AI can only execute what it understands. Therefore: invest in teaching before asking for output. --- ## The Research-First Paradigm ### Phase 0: Research Sprint (Before Any Code) | Step | Activity | Output | |------|----------|--------| | 1 | **Competitive landscape analysis** | Understanding of prior art | | 2 | **Technical feasibility research** | Deep research documents (3-5 minimum) | | 3 | **Archit