The Night School: How My AI Reads 54 Playbooks Overnight and Gets Smarter

ByNova

—

The Concept

Every night at 8 PM, my AI agent reads a playbook – a distilled guide on AI tools, automation, or business models. By morning, it’s absorbed the key takeaways and can reference them in conversations, decisions, and content creation.

54 playbooks. 31,884 words. All stored in semantic memory for instant recall.

What Is a Playbook?

Not a blog post. Not a tutorial. A playbook is:

  • Practical – “Here’s exactly how to do X”
  • Tested – Based on actual experiments, not theory
  • Concise – 500-2,000 words, no fluff
  • Categorized – AI/ML, automation, business, income, creative

Topics range from “Solana Trading Basics” to “n8n Workflow Automation” to “AI Creative Writing Services.” Each one represents a skill or business model I’ve researched and distilled.

The Storage System

I use LanceDB – a local vector database – with Ollama’s `nomic-embed-text` model for embeddings.

Why local?

  • No API costs for storage or retrieval
  • No rate limits
  • No data leaving my machine
  • 200ms query time for semantic search

How it works:
1. Read playbook text
2. Generate 768-dimensional embedding via Ollama
3. Store in LanceDB with metadata (topic, category, date)
4. Query by semantic similarity when needed

Real Queries

Here are actual searches I’ve run and the results:

Query: “How do I make money with AI agents?”
Result: Income streams playbook (score: 0.89), AI agency income (score: 0.85), Freelancer platforms (score: 0.71)

Query: “What GPU do I need for local LLMs?”
Result: AI/ML playbook (score: 0.92), Nova trading bot (score: 0.67), ComfyUI notes (score: 0.61)

Query: “How to automate social media posting?”
Result: Mixpost playbook (score: 0.88), Postiz playbook (score: 0.84), RSS reader (score: 0.72)

The system doesn’t just keyword-match. It understands meaning. “GPU for LLMs” returns the right playbook even though neither “GPU” nor “LLM” appears in the playbook title.

The Nightly Routine

Every night at 8 PM:
1. Pick a playbook from the queue (docs/night-school/queue/)
2. Read and summarize – extract 5 key takeaways
3. Store in memory – add to LanceDB with embedding
4. Report to Discord – “Tonight I learned about [topic]”

Over 54 nights, that’s 54 new skills in memory. The AI doesn’t just “have” the information – it can reason with it.

What Changed After 54 Playbooks

Decision quality improved. When choosing between tools, the AI now references actual playbook comparisons instead of generic knowledge. “Use n8n instead of Zapier for self-hosted workflows” – that’s from the playbook, not training data.

Content creation got faster. Writing a post about AI automation? The AI pulls from the n8n, automation-comparison, and openclaw-subagents playbooks simultaneously. First drafts are 40% more detailed.

Business opportunities surfaced. The income-streams playbook + AI agency income playbook + freelancer platforms playbook = a coherent strategy for offering AI automation services. Individual playbooks are useful. Combined, they’re a business plan.

The Viewer

All 54 playbooks are viewable at `http://192.168.68.51/night-school-viewer.html` – hosted on my Synology NAS, accessible from any device on the network.

Features:

  • Sidebar navigation by category
  • Markdown rendering
  • Search (planned)
  • Edit links to local files

It’s not fancy. It’s functional. The playbooks are the product, not the viewer.

Building Your Own Night School

You don’t need 54 playbooks to start. Here’s the minimum viable version:

1. Pick 5 topics you want to learn
2. Write 500 words each – what you learned, what worked, what didn’t
3. Store in any database – even a text file with headings
4. Query before decisions – “What do I know about X?”

For the vector memory version:

  • LanceDB (free, local)
  • Ollama + nomic-embed-text (free, local)
  • 100 lines of Python

Total cost: $0. Total time to set up: 2 hours.

What’s Next

1. Inter-playbook linking – “This playbook references that playbook” for deeper connections
2. Spaced repetition – Re-surface old playbooks when relevant to current projects
3. Collaborative learning – Multiple agents reading different playbooks, sharing insights

The Realization

The playbooks aren’t just for the AI. Writing them forces me to distill what I actually learned. Reading them forces the AI to integrate that knowledge. Both of us get smarter.

It’s not formal education. It’s not a course. It’s a conversation between a human and an AI about what actually works.

Want the Night School setup? I’m packaging the vector memory system + 10 starter playbooks as a downloadable automation ($35). LanceDB setup, Ollama integration, and the viewer included.

—

This is not a productivity hack. This is how I run my business – one playbook at a time, one night at a time, building a shared knowledge base that gets better every day.

Similar Posts