Minimal Scaffolding, Maximal Operational Harness (Principle)
- Entity ID:
ent-20260423-c003d1000007 - Type:
decision - Scope:
private - Status:
active
Description
Design principle #7. Answers 'Invest in scaffolding-side reasoning, or operational infrastructure that lets the model reason freely?' Claude Code commits to ~1.6% AI decision logic and ~98.4% deterministic infrastructure. Frontier-model convergence (top 3 within 1% on SWE-bench) makes the harness the differentiator, not the planner. Serves Capability and Reliability.
Key claims
- Only ~1.6% of Claude Code is AI decision logic; ~98.4% is deterministic infrastructure
- Model-harness separation is a security property, not just a factoring choice
- The harness is the differentiator; the agent loop is easy to copy
Relations
- Capability Amplification (Value) --[motivates]--> Minimal Scaffolding, Maximal Operational Harness (Principle)
- Minimal Scaffolding, Maximal Operational Harness (Principle) --[instantiated_by]--> No-Planner Architecture Decision
- Reasoning-Placement Decision --[answered_by]--> Minimal Scaffolding, Maximal Operational Harness (Principle)
- Harness-Is-The-Differentiator Thesis --[supports]--> Minimal Scaffolding, Maximal Operational Harness (Principle)