How to use Compass
A step-by-step walkthrough of the full workflow — from first thought to refined insight.
Compass transforms messy thoughts into something measurable.
Most thinking tools ask you to organize what you already understand. Compass does the opposite — it takes an idea you can't fully articulate yet and finds the structure hiding inside it.
Every system — a business, a disease, a relationship, an economy — is made of things that constrain each other across scale. Compass maps those constraints, scores them, and tells you what zone your system is operating in. That's not philosophy — it's measurement.
A project is a focused investigation of a single system or question. Think of it like a lab notebook — not a chat, not a conversation, but a structured inquiry with a beginning, a thesis, and a set of findings.
Don't try to be precise yet. Just describe what you're thinking about. Examples:
Compass listens for the relational grammar underneath your words — not the topic, but how the parts constrain each other.
Compass will ask probing questions and refine your model across a few exchanges. After 2–5 turns it will generate an analysis — a structured map of your system with scored relationships, zone classification, and real-world anchors.
You can keep talking after the analysis is generated. The analysis is a hypothesis — not the end of the conversation.
Compass acts as a guide. It translates your natural language into the framework, asks grounding questions, uses metaphor as a structural tool, and generates human-readable zone explanations. Best for exploring new ideas.
Compass becomes a peer-level co-architect. It speaks formally — Observer Position, Validity Gate, link derivation, measurement roadmaps. Best for researchers or engineers formalizing a model they already understand.
When you've made meaningful progress, save your work using the Save Project button in the top bar of Hierarchics AI. Give the project a short, descriptive name.
Before saving, ask the Librarian to synthesize your project. The Librarian reads the full conversation and produces a structured Markdown document — a re-entry card containing:
- ›Observer Position declared in the session
- ›Key structural findings
- ›Open questions and unresolved tensions
- ›Diagnostic metrics (zone, β values, CC_avg)
- ›A suggested re-entry prompt for the next project
You can also download the synthesis as a .md file.
Compass is designed to be used in cycles. Each project informs the next. The synthesis document is not a final report — it's a seed for the next project.
Hierarchics AI isn't another chatbot. It's a reasoning injection you can drop into any AI you already use.
Each session gives you a structured, hierarchical breakdown of your system — scored relationships, zone classification, and a precise Observer Position. Paste that output into ChatGPT, Claude, Gemini, or any other model, and it becomes noticeably sharper on that topic. Not because you gave it more information, but because you gave it better structure.
It's not about replacing your current AI — it's about making it think better when it matters.
Describe your system in Compass. Let it extract the relational grammar and generate an analysis.
Ask the Librarian to produce a structured re-entry card — a compact, high-density summary of your system.
Paste the synthesis into any other model as context. It now has a precise structural map to reason from.
Synthesis is most valuable when the conversation is coherent, focused, and structurally rich. The Compressibility (CMPRS) indicator in Hierarchics AI tells you how well your current conversation will compress. Here's what makes the difference:
One clear domain. Consistent Observer Position. At least one formal analysis. Moderate-length messages with relational claims — not walls of text. Synthesis will be dense and useful.
Some structure but also some noise. May have a clear domain but no formal analysis, or short exchanges without explicit Observer Position. Synthesis will capture the gist but may miss nuance.
Fragmented topics, very long bloated messages, or incoherent structure. No clear Observer Position. Synthesis may be vague or unreliable — consider refining in Compass first.
Avoid extremely long messages (over 2000 characters) and rapidly switching topics. Each message should advance the relational grammar, not just explore tangentially.
Even in Clarity mode, making your Domain, Granularity, and Intent explicit helps the Librarian produce a sharper synthesis. Ask Compass: "Can we declare the Observer Position for this system?"
Don't synthesize at the exact moment you feel stuck — wait until you've reached a natural pause where the relational claims feel stable. A premature synthesis captures a moving target; a well-timed one becomes a durable re-entry card.
Long projects accumulate context — every message adds to the token count Compass must process on the next call. While a session is active, Compass caches the conversation so each new message only processes the recent tail. But that cache expires after about an hour of inactivity. When you pick up a long project after the cache has gone cold, the entire conversation history must be re-loaded at once — which can trigger a dramatic spike in credit usage on that first call.
This is where synthesis pays for itself. A synthesis compresses the full conversation into a small, high-density document. When you resume a cold project, Compass offers a Resume from Synthesis option: instead of re-ingesting the full raw history, it loads the cached synthesis plus your last few messages — keeping the structural understanding without the bloat.
Before stepping away from a long project, hit Synthesize with Librarian and save. This creates the compressed re-entry card that makes a cheap resume possible later.
When you reopen a large project that's been idle for over an hour, Compass warns you about the credit cost and offers two paths. Choose Resume from Synthesis to load the synthesis as cached context plus your recent messages — a fraction of the token cost of a full reload. A green SYNTH RESUME badge confirms the mode is active.
Once you've resumed from synthesis, the synthesis document is cached for the rest of the session — so subsequent calls stay cheap even as the conversation grows. You keep the full structural context at a fraction of the credit cost.
The Library is your project archive. From there you can:
Resume a saved project and pick up the conversation where you left off.
Read the Librarian's re-entry document for any project that has been synthesized.
Plot multiple projects on the Signature Maps to compare their structural zones.
The Workflow Export turns any analysis into a portable .mmd file — a Mermaid flowchart you can edit, share, or paste into any Mermaid-compatible tool (GitHub, Notion, Obsidian, mermaid.live, etc.).
.mmd format is plain text. Open it in any editor, tweak nodes or edges, and re-render it in mermaid.live or any Markdown tool that supports Mermaid.