Coscientist
GitHub

Coscientist

An intellectual companion system for knowledge production in the post-AI era

Coscientist is an open-source service inspired by Isaac Asimov's fictional Encyclopedia Galactica. It is meant to be more than a knowledge ark of information: an intellectual companion that performs contemplation labor alongside the human Operator. In the post-AI era, its goal is to support un-brain-rotting: restoring verification and agency.

It is built by Sunghyun Cho. The origin story is straightforward: preservation matters, but preservation alone does not produce understanding (Tokens ≠ Knowledge). Coscientist is designed to make knowledge updatable so contradictions, counterexamples, and definition drift become visible and actionable instead of being smoothed away by fluent summaries.

The first principle is sovereignty: knowledge authority must remain with the Operator. Guarding against Encyclopedia Meltdown is not an afterthought; it is a design constraint. AI can do contemplation labor, but it should not seize the initiative of writing and declaring.

This is why Coscientist goes beyond Q&A-style RAG. It maintains a Dialectical Graph that stores argumentative acts rather than raw document text, so every claim can be tied to evidence spans, sources, and explicit relations and Dialectical Graph Edges.

Exploration is adversarial by default. Strategies like counterexample-first search, contradiction isolation via minimum cut, and explicit definition drift detection apply pressure early so knowledge updates by rebuttal instead of drifting by repetition.

At the protocol level, the system uses procedures like the Multi-AI Consensus Protocol and knowledge synthesis to make documents collide rather than merely summarize. Because it uses LLMs, Coscientist can perform cross-linguistic synthesis: reading and synthesizing literature in any language, enabling a truly multilingual knowledge mesh. The intended end state is an ecosystem of arguments: AI handles search and restructuring, while the Operator supervises structure and verification so the resulting narrative remains traceable.

Reference: https://cho.sh/r/6975BA (Opens in a New Tab).

68 Notes Link Here

Edit on GitHub (Opens in a New Tab)

All Notes

147 Notes

  • -Across the Sprachraums
  • -Active Recall
  • -AI
  • -AI Slop
  • -AI-Induced Illusions of Competence
  • -Argumentative Act
  • -Argumentative Relations
  • -As We May Think
  • -Assumption
  • -Attack
  • -Bilingual Cognition
  • -Branched Resolution Map
  • -Claim
  • -Claim Lifecycle
  • -Claim Status Taxonomy
  • -Cognitive Agency Preservation
  • -Cognitive Exoskeleton
  • -Cognitive Sovereignty
  • -Confidence
  • -Contemplation Labor
  • -Contention
  • -Contention as Memorable Anchor
  • -Correction vs Drift
  • 01Coscientist (Currently Open at Position 1)
  • -Counterexample
  • -Counterexample-First Search
  • -Creating Next-gen Digital Brains
  • -Cross-Linguistic Synthesis
  • -Dark Night of the Soul
  • -Definition Drift
  • -Desirable Difficulty in Verification
  • -Deskilling Through AI Delegation
  • -Dialectical Graph
  • -Dialectical Graph Edges
  • -Dialectical Graph Nodes
  • -Dialectical Interleaving
  • -Digital Brain
  • -Digital Garden
  • -Digital Jungle
  • -Document Collision
  • -Drift Phenomena
  • -Encyclopedia Galactica
  • -Encyclopedia Meltdown
  • -Environmental Drift
  • -Epistemic Protocol Layer
  • -Evidence Independence
  • -Evidence Span
  • -Exploration Mechanisms
  • -Exploration Strategies
  • -Extracranial
  • -Federated Knowledge Network
  • -Fluency Trap
  • -Forgetting Curve
  • -Foundation Fiction
  • -Friction as Enemy
  • -From Memex to Dialectical Graph
  • -From Preservation to Capability
  • -Galactic Empire
  • -GitHub for Scientists
  • -Graph as Meltdown Defense
  • -Graph Components
  • -Graph-Based Spaced Repetition
  • -Hallucination
  • -Hari Seldon
  • -Human Agency in AI
  • -Illusions of Competence
  • -Incompatibility Taxonomy
  • -Inference Layer
  • -Institutional Brain Rot
  • -Intellectual Companion
  • -Inter-Sprachraum Communication
  • -Interleaving
  • -Isaac Asimov
  • -Issue Node
  • -Knowledge Ark
  • -Knowledge Constitution
  • -Knowledge Failure Modes
  • -Knowledge Synthesis
  • -Knowledge System Layers
  • -Language-Agnostic Indexing
  • -Learning Science Principles
  • -LLM
  • -Low-Background Steel
  • -Meaning Loss
  • -Memex
  • -Meta-learning
  • -Method
  • -Method-Conclusion Coupling
  • -Minimum Contradiction Set
  • -Minimum Cut
  • -Model Collapse
  • -Monolith as Interface Metaphor
  • -Multi-AI Consensus Protocol
  • -Multilingual Knowledge Mesh
  • -Multilingual Memex
  • -Mystery and Minimalism
  • -Narrative Layer
  • -Natural Science Engineer
  • -Nonstationarity
  • -Normalized Proposition
  • -Operator
  • -Personal Knowledge Evolution
  • -Personal to Institutional Knowledge
  • -Pre-Contamination Resource
  • -Pre-LLM Text
  • -Project Aldehyde
  • -Project PIRI
  • -Provenance
  • -Psychohistory
  • -RAG
  • -RAG Limitations
  • -Rebuttal-First Search
  • -Relation Typing vs Similarity
  • -Replication Path Separation
  • -Responsibility Line
  • -Retrieval Practice
  • -Scapa Flow
  • -ScienceOps
  • -Scope
  • -Second Brain
  • -Seldon Plan
  • -Semantic Drift
  • -Signal Without Explanation
  • -Source
  • -Spaced Repetition
  • -Spacing Effect
  • -Sprachraum
  • -Status Transition Rules
  • -Sunghyun Cho
  • -Superbrain
  • -Synthesis Mechanisms
  • -System Drift
  • -The Monolith
  • -Tokens ≠ Knowledge
  • -Traceability
  • -Training Data Contamination
  • -Translation Fidelity
  • -Translation Nuance Loss
  • -Triple Separation
  • -Un-Brain-Rotting
  • -Unanimity Requirement
  • -Undercut
  • -Vannevar Bush
  • -Verification
  • -Verification as Retrieval Practice
  • -Verification System
  • -Zero-Trust Ingestion