knobs

🎛️ A modular control system for AI-generated nonfiction books
Created by Travis Eric

---

knobs is a strategic control layer for nonfiction book generation. It injects intelligence, depth, and variation into AI-generated content — all while staying deterministic, parallel-safe, and fully phase-aware.

Rather than hardcoding one prompt format, knobs dynamically adjusts each subtopic's depth, specificity, structure, and novelty using modular control switches.

These switches — or knobs — are controlled from a single YAML config file and determine:

• Whether a subtopic should be contrarian or aligned
• Whether it needs a proof point or expert quote
• Whether it should include troubleshooting or failure-modes
• Whether it should be scaled deeper based on phase
• Whether to inject unexpected cross-domain examples

---

🚀 What It Does

• build_knobs(...): decides which knobs to activate for each subtopic
• get_prompt_snippet(...): generates prompt fragments based on those knobs
• depth_multiplier: scales token count and complexity by chapter phase
• wildcard_domain: surprises the model with analogies from wild domains
• All knobs are stateless, reproducible, and independent per subtopic

---

🧠 Why It Matters

Traditional AI generators work linearly:

“Write an outline. Then write a chapter. Then revise.”

knobs is different.

I built it like a system — not a sequence.

• Subtopics don’t need memory of what came before.
• Direction emerges from structure, not serialization.
• Surprise, depth, and precision are shaped per subtopic using deterministic randomness.
• It’s designed to run in parallel across 1,000 pieces — and still sound like a coherent book.

Every subtopic is independently intelligent, but globally aligned.

---

🧩 Use Cases

• AI nonfiction book generators (e.g. Teneo)
• Phase-aware educational content
• Longform idea distillation
• Multi-threaded prompt workflows

It helps inject:

• Strategy
• Variety
• Realism
• Surprise
• Abstraction layering

All with one config file and a prompt adapter.

---

🧪 Example Output

{
  "angle": "contrarian",
  "need_proof": true,
  "troubleshooting_on": false,
  "strategic_on": true,
  "obstacle_on": false,
  "wildcard_domain": "negotiation",
  "depth_multiplier": 1.77
}


This knobset would trigger:
- A strategic lens
- A counter-argument framing
- An example from negotiation
- Longer, deeper content output

⚙️ Config via knobs.yml
yaml
Copy
Edit
angle_prob: 0.05
need_proof_prob: 0.10
wildcard_domains:
  - negotiation
  - parenting
  - jazz improvisation
  - museum curation
  - esports coaching
depth_by_phase:
  foundation: 1.0
  development: 1.18
  mastery: 1.77
Change the knobs → change how your content feels.

📁 Files
File	Purpose
knobs.py	Core logic (build_knobs, get_prompt_snippet)
knobs.yml	Config file for probabilities and domains
prompt_snippets.md	All guidance blocks knobs can inject
example_knob_output.json	Sample knobsets for analysis

🧠 System Thinking vs. Linear Generation
Most AI writing systems are linear. They move from start to finish like an author: outline → intro → body → polish.

I didn’t build it that way.

This framework is parallel-safe, prompt-independent, and depth-scalable.
Every subtopic is treated like a node in a distributed system — not a line in a narrative.

That’s why this works at scale. That’s why it’s flexible. That’s why it feels intentional, not templated.

You can’t build emergent strategy with linear prompting.
You need knobs.

🛠 Integration
Use knobs with any generator that:

Builds structured prompts for subtopics

Accepts per-subtopic guidance

Supports token scaling or modular prompt logic

Ideal for:

Longform AI generation

Instructional books

Modular course builders

Content automation systems

📘 License
MIT — fork it, remix it, build with it.
If you make something cool with it, link back or say hi.

Built by Travis Eric
traviseric.com · @traviseric_