← All Lessons
Lesson 10

The Hive

A daily AI scout that watches all your projects, scores their health, and tells you where to focus.

You have too many projects. Some are thriving, some are stale, some are dying and you don't even know it. The Hive crawls all of them daily, scores their health, and generates a "waggle dance" report telling you exactly where your attention should go.

What you're building today

An automated ecosystem dashboard. A Python script crawls your project folders, checks git activity, file freshness, and project signals. Then Claude analyzes the data and generates a health report. The result is a single-file HTML dashboard you can open every morning.

Step 1: The concept

In a real beehive, scout bees return from the field and do a "waggle dance" to tell the colony where the best nectar is. Your Hive does the same thing for your projects.

42+ project folders on your machine
    ↓
hive.py (Python collector)
    ↓  crawls each folder
    ↓  checks: last commit, file count, README, CLAUDE.md
    ↓  scores: nectar (value), freshness, tension
    ↓  writes JSON snapshot
    ↓
analyze.sh (orchestrator)
    ↓  feeds snapshot to claude -p
    ↓  Claude generates waggle dance report
    ↓  renders single-file HTML dashboard
    ↓
dashboard.html (open in browser)
    Top 5 projects to focus on today
    Health scores across your entire portfolio

Step 2: The signal collector

The collector walks through your project folders and gathers signals automatically:

# hive.py — the signal collector
import os, json, subprocess
from datetime import datetime
from pathlib import Path

PROJECTS_DIR = Path.home() / "strategy" / "projects"
SNAPSHOT_DIR = Path("snapshots")

def scan_project(path):
    """Score a single project's health signals."""
    signals = {
        "name": path.name,
        "path": str(path),
        "has_readme": (path / "README.md").exists(),
        "has_claude_md": (path / "CLAUDE.md").exists(),
        "file_count": sum(1 for _ in path.rglob("*") if _.is_file()),
    }

    # Check git activity
    try:
        last_commit = subprocess.check_output(
            ["git", "log", "-1", "--format=%ci"],
            cwd=path, stderr=subprocess.DEVNULL
        ).decode().strip()
        signals["last_commit"] = last_commit
    except:
        signals["last_commit"] = None

    return signals

# Scan all projects
projects = []
for p in sorted(PROJECTS_DIR.iterdir()):
    if p.is_dir() and not p.name.startswith("."):
        projects.append(scan_project(p))

# Save snapshot
SNAPSHOT_DIR.mkdir(exist_ok=True)
date = datetime.now().strftime("%Y-%m-%d")
snapshot_path = SNAPSHOT_DIR / f"{date}.json"
snapshot_path.write_text(json.dumps(projects, indent=2))
print(f"Snapshot: {snapshot_path} ({len(projects)} projects)")

Step 3: The AI analyzer

This is where Claude turns raw data into judgment. The orchestrator script feeds the snapshot to Claude and asks for a waggle dance.

#!/bin/bash
# analyze.sh — collect + analyze + render

# Step 1: Collect signals
python3 hive.py

# Step 2: Feed to Claude for analysis
DATE=$(date +%Y-%m-%d)
SNAPSHOT="snapshots/$DATE.json"

claude -p "
You are a project health analyzer. Here is a JSON snapshot
of all projects in my portfolio:

$(cat $SNAPSHOT)

Generate a waggle dance report:
1. Score each project: THRIVING / ACTIVE / STALE / DORMANT
2. Rank the top 5 projects I should focus on today
3. Flag any projects that haven't been touched in 14+ days
4. Note any projects missing CLAUDE.md (they should have one)

Be direct. No filler. Format as markdown.
" --dangerously-skip-permissions --max-turns 3 > "reports/$DATE.md"

echo "Report: reports/$DATE.md"

Step 4: The dashboard

The report is useful as markdown, but it's better as a visual dashboard you can open every morning.

The dashboard is a single HTML file with the data embedded. No server, no database, no dependencies. Just open it in your browser. It shows project health as colored indicators: green for thriving, amber for active, red for stale, gray for dormant.

Step 5: Run it daily

# Add to your crontab (runs at 8am every day)
crontab -e

# Add this line:
0 8 * * * cd ~/strategy/projects/hive && ./analyze.sh

Or just run it manually when you sit down in the morning. The point is: one command gives you a complete picture of your entire portfolio.

Step 6: What you learn from it

The Hive doesn't just show you what's active. It shows you patterns:

The daily snapshot creates a time series. After a month, you can diff today vs 30 days ago and see exactly how your portfolio shifted. That's not a todo list. That's strategic intelligence.

What you just built

Homework

Set up the collector on your strategy folder. Run it once manually. See how Claude scores your projects. You'll probably be surprised which ones it flags as stale. That's the point. The Hive sees what you don't have time to look at.