Continuous Learning v2 FAQs

Question 1

How does Continuous Learning v2 differ from standard AI memory?

Accepted Answer

Unlike standard context windows, this skill uses deterministic hooks to analyze every action and create permanent, weighted behavioral rules (instincts) that evolve into reusable skills and commands.

Question 2

Is my private code or session data shared externally?

Accepted Answer

No. All observations, instincts, and evolved skills are stored locally on your machine in the ~/.claude/homunculus directory, ensuring complete privacy and data sovereignty.

Question 3

How does the evolution process work?

Accepted Answer

The system clusters related instincts over time. When a specific pattern is observed frequently enough to hit the 'cluster threshold', the /evolve command generates a dedicated skill or agent to handle that specific domain.

Question 4

What is the 'confidence scoring' system?

Accepted Answer

Every learned instinct is assigned a score from 0.3 to 0.9. Low scores are treated as tentative suggestions, while scores of 0.7 and above become core behaviors that Claude applies automatically.

Question 5

Does it slow down my Claude Code performance?

Accepted Answer

No. Observations are handled via lightweight hooks and the analysis is performed by a background agent (Haiku) to ensure your active development session remains fast and responsive.

Continuous Learning v2

Key Features

Use Cases

Continuous Learning v2

Key Features

Use Cases