Skip to content

QUICK-START.md

Latest commit

 

History

History
162 lines (102 loc) · 3.63 KB

QUICK-START.md

File metadata and controls

162 lines (102 loc) · 3.63 KB

Quick Start Guide

Get LC-OS running in 30 minutes.

What You Need

  • An AI assistant (ChatGPT, Claude, Gemini, or similar)
  • A place to store files (local folder, cloud drive, or GitHub)
  • A project that will span multiple sessions

Step 1: Create Your Running Document (10 minutes)

Copy the template from minimal/running-document-template.md

Fill in:

  • Project name
  • What you are trying to accomplish
  • Any constraints or rules that apply
  • Key decisions already made

This is your shared memory. Everything important goes here.

Step 2: Create Your Canonical Numbers File (5 minutes)

Copy the template from minimal/canonical-numbers-template.md

Add any numbers that matter:

  • Budgets
  • Deadlines
  • Targets
  • Rates or prices

If a number is not in this file, it is not authoritative.

Step 3: Start Your First Session (5 minutes)

Begin your AI session with:

Here is our Running Document for this project. Please read it before we begin.

[Paste or attach your Running Document]

Key rules:
- Reference this document for context and decisions
- All numbers come from the Canonical Numbers file only
- If something is unclear, ask one clarifying question then proceed

Step 4: Work Normally

Do your work as usual. The difference:

  • When you make a decision, note it in the Running Document
  • When you confirm a number, add it to Canonical Numbers
  • When the AI forgets context, point it back to the Running Document

Step 5: When Something Goes Wrong

Do not push through. Use this sequence:

  1. Stop - Pause the current task
  2. Diagnose - What went wrong and why?
  3. Rollback - Return to the last correct state
  4. Note - Log the failure and update documents

Then continue from the stable state.

Step 6: End of Session

Before closing:

  • Update the Running Document with any new decisions
  • Add any new numbers to Canonical Numbers
  • Note any failures in your Failure Log

This takes 2-3 minutes and saves significant time in future sessions.

What You Will Notice

After a few sessions:

  • Less time re-explaining context
  • Fewer numerical inconsistencies
  • Clearer sense of what has been decided
  • Faster recovery when things go wrong

These gains compound over time.

Next Steps

If minimal is working: Continue with minimal. Do not add complexity you do not need.

If you want more structure: Explore the full/ folder for:

  • Repair Protocol (structured recovery)
  • Stability Pings (alignment checks)
  • Challenge Protocol (productive disagreement)
  • Affective Governance (tone management)

If failures are frequent: Start using the Failure Log template to track patterns.

Common Questions

Which AI does this work with? Any conversational AI: ChatGPT, Claude, Gemini, and others. The principles are model-agnostic.

Do I need to share the full Running Document every session? Yes, at the start of each new session. The AI does not remember between sessions by default.

What if my project is simple? You might not need LC-OS. It is designed for work that spans multiple sessions where consistency matters.

Can I modify the templates? Yes. Adapt them to your needs. The structure is a starting point, not a rigid requirement.

One-Page Summary

BEFORE EACH SESSION
- Share Running Document with AI

DURING SESSION  
- Reference documents for context
- Note decisions as they are made
- Numbers only from Canonical Numbers

WHEN THINGS BREAK
- Stop → Diagnose → Rollback → Note

END OF SESSION
- Update Running Document
- Log any failures

That is LC-OS at its simplest.

Start here. Add structure only when you need it.