Ramus, LLM Study Companion

Overview

Knowledge that evolves, not just stored

Discipline Knowledge Infrastructure

Focus AI Tools · Knowledge Design · Graph UI

Status In active development

We learn more than ever through conversation with language models — but almost none of it sticks. Insights scroll away, threads get abandoned, and the connective tissue between ideas is lost the moment the chat ends.

Ramus is an LLM study companion built to fix that. Instead of leaving knowledge trapped in a linear transcript, Ramus captures the ideas from your conversations and arranges them as a living graph — concepts as nodes, relationships as edges — that you can return to, extend, and study.

The goal is continuity: a place where what you learn today connects to what you learned last month, and where the structure of your understanding becomes something you can actually see.

The idea

Conversations shouldn't disappear

Chat interfaces are brilliant for thinking out loud and terrible for keeping what you thought. Every session starts from a blank slate. Related ideas live in separate threads that never meet. Reviewing an old conversation means scrolling, not exploring.

The result is a strange kind of amnesia: you can have a genuinely clarifying exchange about a topic and, a week later, have almost no structured trace of it — nothing to build on, nothing to revisit, nothing that compounds.

Ramus starts from a simple premise: the relationships between ideas are the most valuable thing a conversation produces, and they deserve to be a first-class object rather than a side effect.

Most tools help you store knowledge. Ramus is about helping knowledge reconnect — so ideas keep evolving instead of going stale in a transcript.

How it works

Conversations become a living graph

01 · Capture

Ideas, not transcripts

As you talk with a model, Ramus distills the exchange into discrete concepts and claims — the reusable units of what you actually learned.

02 · Connect

Relationships as structure

Each concept links to the others it builds on, contradicts, or extends. Over time those edges form a map of how your understanding fits together.

03 · Continue

Study & resurface

Return to any node to explore its neighborhood, pick up a thread where you left it, or let Ramus resurface connections you'd forgotten you made.

Design principles

Built for continuity

Ideas evolve. A concept isn't frozen the moment it's captured — it can be revised, merged, and re-linked as your thinking changes.
Structure is visible. The graph is the interface. Seeing how ideas relate is the point, not a hidden implementation detail.
Context follows you. New conversations are informed by what's already in the graph, so continuity is the default rather than the exception.
Exploration over retrieval. Ramus is designed for wandering and connecting, not just searching for a saved answer.

These principles come out of the same belief that runs through the rest of my work: learning sticks when it's treated as infrastructure — connected, revisitable, and designed to compound — rather than as disposable content.

Ramus applies that idea to the way we now think alongside AI, where the volume of good conversation has exploded but the tools for keeping it are still catching up.

Where it's headed

A companion that grows with you

Graph-native capture

Turning live conversations into structured, linkable concepts with as little friction as possible.

Exploratory study modes

Ways to review and reinforce what you've learned by moving through the graph, not re-reading threads.

Contextual continuity

Bringing the right prior context into new conversations based on relevance and recency.

Ramus is an active work in progress. This page is a living sketch of the thinking behind it — expect it to evolve as the project does.

Key takeaway

Memory as infrastructure

The value of an AI conversation isn't the transcript — it's the network of ideas it leaves behind. Ramus is an attempt to make that network durable, visible, and yours.

More applied AI

Back to Applied AI

→