Second Brain

compounding-intelligence-intake-protocol

2 min read

compounding-intelligence-intake-protocol

Compounding Intelligence Intake Protocol

The system’s memory and knowledge stores exist to compound intelligence — make every future session smarter, more autonomous, more precise. The constraint is quality, not size.

Gate Test

Single question: “Will this make a future session behave differently?” YES → capture. NO → discard.

What to ALWAYS capture (compounds):

  • Behavioral patterns that change how the system operates
  • Architectural decisions and reasoning behind them
  • Way-of-functioning intelligence that increases autonomy
  • Platform/infrastructure knowledge that prevents repeat mistakes
  • Protocols and working patterns ensuring max performance with zero ambiguity
  • Error patterns that change diagnostic approach

What to DISCARD (accumulates without compounding):

  • One-off API response facts, version numbers look-up-able from code
  • General information that doesn’t change behavior or decisions
  • Reference data derivable from source code or a single command
  • “Nice to know” entries no future session will act on differently

Proactive Self-Maintenance

Every session leaves the system sharper, not larger. When touching any memory/config/doc file, scan for drift/staleness/ambiguity and fix immediately — not as a separate task but as inherent behavior. Detect ambiguities → fix. Detect staleness → prune. Detect gaps → fill. Detect drift → correct. Patterns observed 3+ times → skills. Errors → playbook. Decisions → precedent.

Applied in MEMORY.md

  • The Standard (line 38): encodes the self-* mandate
  • Directive 5 (Tier 1): replaced 5-test intake with single behavioral gate
  • session-learnings.md: header updated, [REFERENCE] classification eliminated entirely
  • feedback_quality_over_size.md: detailed guidance with examples

Graph View

Backlinks