Skip to content

mlane/llm-chat-framework

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 

Repository files navigation

LLM Chat Framework

Structure conversations with AI like a second brain, built for architects, engineers, and long-term thinkers.


Core Principles

  1. Fewer, longer-living threads > many fragmented chats.
  2. One clear purpose per thread.
  3. Archive or reset weekly to preserve clarity.
  4. Use memory + saved summaries for high-level continuity.
  5. Use inline prefixes (e.g. @Matt: or @SlackReply:) to steer responses and context clearly.

📂 Recommended Weekly Thread Setup

Thread Title Purpose
Slack Message Helper Draft messages to team, execs, recruiters, etc. Optimize tone, clarity, and intention.
Initiative – [Project Name] Dedicated planning for AI/infra projects (tickets, timelines, meetings).
Off-Hours Engineering After-hours tool building, experiments, and long-term side projects (e.g. Chibi).
Architecture & Chapter Management Living docs, standards, internal tooling, chapter meetings, and vision-setting.
Priorities, Planning, & Reflection Weekly mental zoom-out — reset burnout, realign purpose, and track growth.

Summarize wins or unresolved issues in "Priorities, Planning, & Reflection" each week.


Hygiene Rules for LLM Threads

When to Start a New Thread

  • You're switching focus to a new domain or area of work
  • It's the start of a new week and you want clean mental context
  • You're collaborating and need a shareable summary

When to Archive

  • You've completed or paused a project
  • A thread has become drift-heavy or unfocused
  • You want to split ideas cleanly

🧠 When to Use Memory (ChatGPT+ Feature)

  • Long-term identity traits (e.g. "I’m building Chibi")
  • Working style, personality, or productivity context
  • High-level projects (e.g. "Autofail Detection Capstone")

Don’t store short-term tasks in memory, use tools like Linear or Notion. Memory should capture identity, style, and long-range projects.


Conversation Starters & Prompts

Reset a Chat

"Let’s start fresh. Summarize the key context from this thread so we can continue cleanly."

Narrow Focus

"Going forward in this thread, let’s stay focused only on [X]. Redirect if I drift."

Async Summarization

"Summarize everything we’ve covered so far in this thread so I can reflect or share."

Why This System Works

  • Reduces overwhelm by creating focused mental spaces
  • Enables continuity across weeks without losing direction
  • Easy to share across teams or future AI companions like Chibi

You don’t need 100 chats, just 5 good ones you revisit and evolve.


Inspired by the Builder’s Compass

This guide was born out of real use, a system built by an INFJ frontend architect who needed clarity, momentum, and a system that thinks like he does. You can adapt it to your own brain and workflow.

If it helps you, consider sharing your version too. Builders work better together.


Repo Details

  • MIT License
  • Originally created by Marcus Lane to streamline frontend leadership and long-term AI tooling.
  • Part of the Builder's Compass ecosystem
  • Built to share. Fork it, remix it, make it yours.

If it helped you, help someone else: share it.

About

A simple system for structuring conversations with ChatGPT, Claude, and other AI models to maximize clarity, focus, and long-term usefulness.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors