learn-new-things

Learn New Things - Continuous Learning Heartbeat

Autonomous learning loop that periodically expands the knowledge base through research, extraction, and connection discovery. Runs locally using existing skills and sub-agents.

Overview

This heartbeat skill implements a continuous learning cycle:

1. Research - Discover cutting-edge papers and developments
2. Extract - Pull unique insights into Document Insights
3. Connect - Map discoveries to existing knowledge base
4. Commit - Save results to dedicated git branch, return to main
5. Rest - Wait for next cycle

Each cycle is a complete learning session. The heartbeat never "completes" - it continuously learns.

Git Workflow: Each learning session commits to its own branch (learning/YYYY-MM-DD-topic-slug), then returns to main. This keeps main clean while preserving all learning for selective merging.

Dependencies

• Skills: /deep-research, /auto-discovery, /integrate-recent-notes, /refresh-index
• Sub-agents: research-specialist, document-insight-extractor, connection-finder
• Local Brain Search: For semantic search and connection discovery

Usage

/learn-new-things                    # Default: 8-hour interval, auto-select topic
/learn-new-things 4                  # 4-hour interval
/learn-new-things 8 "multi-agent systems"  # Specific topic
/learn-new-things stop               # Stop the learning loop

---

State Tracking

Track learning progress in resources/learn-new-things-log.md:

session_id: YYYY-MM-DD-HHMMSS
last_cycle: 2026-02-18T13:15:00
cycles_completed: 0
topics_researched: []
insights_extracted: 0
connections_discovered: 0
consecutive_errors: 0
phase: "running"  # running | paused | error
branches_created: []  # e.g., ["learning/2026-02-18-embodied-cognition"]

---

STEP 1: Initialize State

Read or create state file:

cat resources/learn-new-things-log.md 2>/dev/null || echo "No existing state"

Parse arguments:

• $ARGUMENTS[0] - Interval in hours (default: 8)
• $ARGUMENTS[1] - Optional topic (default: auto-select)

If argument is "stop", set phase: "paused" and exit.

---

STEP 2: Pre-Cycle Preparation

Ensure Index is Fresh

Check when index was last updated:

ls -la resources/local-brain-search/data/brain.faiss

If older than 24 hours, refresh:

resources/local-brain-search/run_reindex.sh

Check Knowledge Base State

Read current analysis:

head -100 knowledge-base-analysis.md

Note:

• Current note counts
• Identified gaps
• Recent research sessions

---

STEP 3: Topic Selection

If Topic Provided

Use the provided topic from $ARGUMENTS[1].

If Auto-Select (Default)

Select topic based on knowledge base gaps and rotation. Use these strategies:

Strategy A: Gap-Filling Choose from underrepresented domains in knowledge-base-analysis.md:

• Systems thinking & complexity science (12 notes - gap)
• Embodiment & interoception (14 notes - gap)
• Creativity neuroscience
• Memory consolidation
• Collective intelligence

Strategy B: Depth-Building Extend existing strong domains:

• AI agent architectures (latest 2025-2026 developments)
• Neuroscience of decision-making
• Buddhism-neuroscience bridges
• Identity and belief systems

Strategy C: Emerging Trends Research cutting-edge developments:

• Latest AI safety research
• New consciousness research
• Recent dopamine/motivation findings
• Multi-agent coordination

Rotation Logic:

cycle_num = cycles_completed % 3
if cycle_num == 0: Strategy A (gap-filling)
if cycle_num == 1: Strategy B (depth-building)
if cycle_num == 2: Strategy C (emerging)

Document selected topic and rationale.

---

STEP 4: Execute Deep Research

Launch the deep-research skill with selected topic:

Use Task tool with subagent_type='research-specialist':

TOPIC: [Selected topic]

Conduct comprehensive research on [topic] focusing EXCLUSIVELY on the most recent research and developments (2025-2026).

SEARCH STRATEGY:
- Prioritize papers from last 12-18 months
- Search for "2025", "2026", "recent", "latest" in queries
- Check arXiv preprints, major conferences (NeurIPS, ICML, ICLR)
- Look for industry whitepapers and blog posts

OUTPUT REQUIREMENTS:
- 15-25 major papers/developments
- Full citations with DATES
- Key findings and novel contributions
- Save to: resources/[Topic-Slug]-Research-YYYY-MM-DD.md

On Success: Continue to Step 5 On Failure: Log error, increment consecutive_errors, check threshold

---

STEP 5: Extract Insights

Create Session Folder

Format: YYYY-MM-DD [Topic Description]

date '+%Y-%m-%d'
# Create: Brain/Document Insights/YYYY-MM-DD [Topic]/

Launch Document Insight Extractor

Use Task tool with subagent_type='document-insight-extractor':

Extract unique insights from the research report for the knowledge base.

SOURCE DOCUMENT: [Path to research report from Step 4]
SESSION FOLDER: [Session folder name]

EXTRACTION GUIDELINES:
1. Focus on novel insights (paradigm shifts, counter-intuitive findings)
2. Bridge to existing hubs: Consciousness, Dopamine, Decision-Making, Identity, AI Agents, Flow
3. Quality > Quantity: 15-25 high-value insights
4. ALWAYS search for duplicates before creating notes
5. Create changelog in session folder

On Success: Count insights extracted, continue to Step 6 On Failure: Log error, continue to Step 6 (partial success is OK)

---

STEP 6: Discover Connections

Launch Connection Finder

Use Task tool with subagent_type='connection-finder':

Discover connections between newly extracted insights and existing knowledge base.

STARTING POINTS: All notes in session folder: [Session folder path]

CONNECTION DISCOVERY GOALS:
1. Bridge to existing your permanent notes
2. Link to 6 primary thematic hubs
3. Find cross-domain consilience opportunities
4. Similarity thresholds: 0.65-0.85

OUTPUT:
- Connection map for new insights
- Synthesis opportunities identified
- Changelog: CHANGELOG - Connection Discovery Session YYYY-MM-DD.md in Brain/05-Meta/Changelogs/

On Success: Count connections, continue to Step 7 On Failure: Log error, continue to Step 7

---

STEP 7: Update State & Log

Update State File

Write to resources/learn-new-things-log.md:

# Learn New Things - Session Log

**Session ID:** [session_id]
**Last Updated:** [timestamp]
**Phase:** running

## Statistics
- Cycles completed: [N]
- Topics researched: [list]
- Total insights extracted: [N]
- Total connections discovered: [N]
- Consecutive errors: [N]

## Latest Cycle
- **Started:** [timestamp]
- **Topic:** [topic]
- **Research report:** [path]
- **Session folder:** [path]
- **Insights extracted:** [N]
- **Connections found:** [N]
- **Status:** [success/partial/error]

## Cycle History
| Date | Topic | Insights | Connections | Status |
|------|-------|----------|-------------|--------|
| YYYY-MM-DD | [topic] | [N] | [N] | [status] |

Log to Master Changelog

Add entry to Brain/CHANGELOG.md:

## YYYY-MM-DD - Learning Heartbeat Cycle [N]

- **Topic:** [topic]
- **Insights extracted:** [N]
- **Connections discovered:** [N]
- **Session folder:** [[Document Insights/YYYY-MM-DD Topic]]

---

STEP 8: Git Commit & Branch Management

After completing the learning cycle, commit all changes to a dedicated branch, then return to main.

Create Branch Name

Generate branch name from topic and date:

# Get current date
DATE=$(date '+%Y-%m-%d')

# Create topic slug (lowercase, hyphens, no special chars)
# Example: "Multi-Agent Systems" → "multi-agent-systems"
TOPIC_SLUG=$(echo "[topic]" | tr '[:upper:]' '[:lower:]' | sed 's/[^a-z0-9]/-/g' | sed 's/--*/-/g' | sed 's/^-//' | sed 's/-$//')

BRANCH_NAME="learning/${DATE}-${TOPIC_SLUG}"

Ensure Clean State on Main

Before creating the learning branch, ensure we're on main:

cd "$(git rev-parse --show-toplevel)"
git stash --include-untracked -m "Pre-learning stash $(date '+%Y-%m-%d %H:%M')" 2>/dev/null || true
git checkout main
git pull origin main 2>/dev/null || true
git stash pop 2>/dev/null || true

Create and Switch to Learning Branch

git checkout -b "$BRANCH_NAME"

Stage Learning Results

Stage all files created during this cycle:

# Research report
git add "resources/[Topic-Slug]-Research-*.md"

# Document Insights session folder
git add "Brain/Document Insights/[Session-Folder]/"

# Changelogs
git add "Brain/05-Meta/Changelogs/CHANGELOG - *.md"
git add "Brain/CHANGELOG.md"

# State file
git add "resources/learn-new-things-log.md"

# Local Brain Search index updates (if any)
git add "resources/local-brain-search/data/" 2>/dev/null || true

Commit with Descriptive Message

git commit -m "$(cat <<'EOF'
Learning: [Topic] - Cycle [N]

Research & Extraction Session:
- Topic: [topic]
- Papers analyzed: [N]
- Insights extracted: [N]
- Connections discovered: [N]

Session folder: Brain/Document Insights/[Session-Folder]/
Research report: resources/[filename]

Generated by /learn-new-things heartbeat
EOF
)"

Push Branch to Remote

git push -u origin "$BRANCH_NAME"

Create Pull Request

Create a PR for review and selective merging:

gh pr create --title "Learning: [Topic] - Cycle [N]" --body "$(cat <<'EOF'
## Learning Session Summary

**Topic:** [topic]
**Date:** YYYY-MM-DD
**Cycle:** [N]

### Research Results
- Papers analyzed: [N]
- Insights extracted: [N]
- Connections discovered: [N]

### Files Added
- Research report: `resources/[filename]`
- Session folder: `Brain/Document Insights/[Session-Folder]/`
- Changelogs updated

### Key Discoveries
1. [Most significant insight]
2. [Cross-domain connection found]
3. [Synthesis opportunity identified]

### Review Checklist
- [ ] Insights are high quality and non-redundant
- [ ] Connections to existing notes are valid
- [ ] No sensi

---

*Content truncated.*