Historical Note (October 2025): This project was originally named "TrueNames" (inspired by Ursula K. Le Guin's "The Rule of Names"). We rebranded to Mamut Lab in October 2025 to avoid confusion with TrueName.ai, an existing AI company.

The new name honors the Kikinda mammoth ("Kika"), discovered in Kikinda, Serbia in 1996—preserved for 500,000 years, just like research context in our platform.

The Le Guin inspiration and literary heritage remain part of our story, documented below for posterity.

The Mamut Lab idea is built upon decades of academic research and the brilliant work of many researchers. This page acknowledges the foundational ideas, papers, and people that made this concept possible.

Current Status: Mamut Lab is currently in the prototyping and early implementation phase. Core features are being developed based on the research documented here.

Core Principle: All innovation is incremental. The Mamut Lab concept synthesizes and applies existing research—it does not claim to have invented these concepts. Credit belongs to the original researchers.

📖 Original Literary Inspiration: "TrueNames"

Ursula K. Le Guin's "The Rule of Names"

Our original name "TrueNames" was inspired by Ursula K. Le Guin's short story "The Rule of Names" (1964) from the Earthsea series.

The Concept: In Le Guin's Earthsea universe, everything has a "true name" in the Old Speech. To know something's true name is to understand its essential nature and have power over it. Wizards guard their true names carefully, as knowing someone's true name gives you deep insight into—and potential control over—their essence.

Connection to This Project: This philosophy remains central to Mamut Lab: truly understanding code, decisions, and systems rather than accepting AI outputs as mysterious black boxes. Just as knowing a true name in Earthsea means genuine understanding, this platform aims to reveal the "true names" (deep reasoning) behind every AI decision.

Attribution:

  • Author: Ursula K. Le Guin (1929-2018)
  • Story: "The Rule of Names" (1964)
  • Series: Earthsea Cycle
  • Read: Available in the collection "The Wind's Twelve Quarters" and various Earthsea anthologies
  • Learn More: Official Ursula K. Le Guin Website
  • Find the Story: Available through libraries, bookstores, and major retailers

"You must not change one thing, one pebble, one grain of sand, until you know what good and evil will follow on that act. The world is in balance, in Equilibrium. A wizard's power of Changing and Summoning can shake the balance of the world. It is dangerous, that power. It is most perilous. It must follow knowledge, and serve need."

— Ursula K. Le Guin, A Wizard of Earthsea

Why This Matters: Like a wizard must understand the true nature of what they work with, this platform insists that developers understand the true nature of the code and decisions AI systems produce. No shortcuts, no mysteries—only genuine understanding.

🔬 Technical & Academic Foundations

Beyond the literary inspiration for the name, the technical implementation is built on decades of academic research:

🧬 Darwin-Gödel Self-Improvement

Foundational Research

The Darwin Gödel Machine (2025)

  • Authors: Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, Jeff Clune
  • Institution: Sakana AI, University of British Columbia
  • Contribution: First practical implementation of self-improving AI that recursively modifies its own code, achieving 2.5× performance improvements on software engineering tasks
  • Key Innovation: Replaced intractable formal proofs with empirical testing, making Gödel Machine concept practical

Gödel Machine (2003)

  • Author: Jürgen Schmidhuber
  • Institution: IDSIA, Switzerland
  • Contribution: First mathematically rigorous framework for provably optimal self-improving AI
  • Concept: Systems that only modify their code after finding formal proofs that modifications improve performance

Evolution Through Large Models (2022)

  • Authors: OpenAI Research Team
  • Contribution: Using large language models to propose intelligent mutations in evolutionary algorithms

MAP-Elites Algorithm

  • Authors: Jean-Baptiste Mouret, Jeff Clune
  • Contribution: Quality-diversity algorithm maintaining diverse populations rather than single optimal solutions

🧠 Neurosymbolic AI & Knowledge Graphs

Core Research

Neurosymbolic AI Foundations

  • Researchers: Numerous contributions from Artur d'Avila Garcez, Luis Lamb, Tarek Besold, and many others
  • Concept: Combining neural networks (learning from data) with symbolic reasoning (logical inference)
  • Why it matters: Enables explainable AI through hybrid neural-symbolic architectures

Graph Neural Networks (GNN)

  • Researchers: Pioneered by multiple research groups including Kipf & Welling (2017)
  • Contribution: Neural networks that operate on graph structures, enabling reasoning over knowledge graphs

Knowledge Graph Research

  • Foundations: Semantic web research, RDF, OWL ontologies
  • Modern Applications: Google Knowledge Graph, academic knowledge bases

🎓 Continual Learning & Catastrophic Forgetting

Foundational Work

Elastic Weight Consolidation (EWC)

  • Authors: James Kirkpatrick et al., DeepMind
  • Publication: PNAS 2017
  • Contribution: Method for preventing catastrophic forgetting by protecting important neural network parameters

Experience Replay

  • Origins: Reinforcement learning literature
  • Application: Periodically rehearsing previous examples to maintain performance on old tasks

Task-Incremental, Domain-Incremental, Class-Incremental Learning

  • Community: Continual learning research community
  • Taxonomy: Framework for categorizing different types of continual learning scenarios

🧩 Dual-Process Cognition

Psychological Foundations

Thinking, Fast and Slow

  • Author: Daniel Kahneman (Nobel Prize in Economics)
  • Co-researcher: Amos Tversky
  • Concept: Human cognition operates through System 1 (fast, intuitive) and System 2 (slow, deliberative) thinking
  • Application: Routing AI decisions between fast neural inference and slow symbolic reasoning

Flow Theory

  • Author: Mihály Csíkszentmihályi
  • Concept: Optimal performance occurs when challenge matches skill level
  • Application: Adaptive difficulty calibration, skill maintenance protocols

🏗️ Software Architecture Patterns

Event Sourcing & CQRS

Command Query Responsibility Segregation (CQRS)

  • Pioneer: Greg Young
  • Concept: Separate read and write models for optimal performance and scalability

Event Sourcing

  • Community: Domain-Driven Design community, Martin Fowler's influence
  • Concept: Persist all changes as immutable events, enabling time-travel debugging and complete audit trails

Domain-Driven Design (DDD)

Foundational Work

  • Author: Eric Evans
  • Book: "Domain-Driven Design: Tackling Complexity in the Heart of Software" (2003)
  • Concepts: Bounded contexts, ubiquitous language, aggregates

Hexagonal Architecture (Ports & Adapters)

Pattern Origin

  • Author: Alistair Cockburn
  • Concept: Dependency inversion through ports and adapters for testability and flexibility

🔬 AI Safety & Alignment

Research Foundations

AI Alignment Research

  • Researchers: Stuart Russell, Eliezer Yudkowsky, Paul Christiano, and many others
  • Concepts: Value alignment, corrigibility, reward hacking prevention
  • Application: Cascade prevention systems, goal drift detection, human oversight gates

Tripwire Mechanisms

  • Concept: Predefined boundaries that halt execution when crossed
  • Application: Safety constraints preventing runaway optimization

🌐 Distributed Systems & Orchestration

Workflow Orchestration

Temporal.io

  • Origins: Uber's Cadence project
  • Contributors: Maxim Fateev, Samar Abbas, and team
  • Concept: Durable execution for long-running workflows with fault tolerance

Saga Pattern

  • Origin: Database transaction literature
  • Application: Distributed transaction management with compensation

Observability

OpenTelemetry

  • Community: CNCF (Cloud Native Computing Foundation)
  • Contribution: Unified observability framework (traces, metrics, logs)

💾 Data Architecture

Polyglot Persistence

Concept Origins

  • Popularized by: Martin Fowler, Neal Ford
  • Concept: Use different databases for different use cases (graph, document, key-value, etc.)

Multi-Model Databases

  • Example: ArangoDB (graph + document + key-value in one system)
  • Contributors: ArangoDB team and community

📚 Language & Technology Inspirations

Programming Languages

Python

  • Creator: Guido van Rossum
  • Why it matters: Dominant AI/ML ecosystem (PyTorch, TensorFlow, Hugging Face)

Go

  • Creators: Robert Griesemer, Rob Pike, Ken Thompson (Google)
  • Why it matters: Excellent for orchestration and control planes (proven at Stream, Discord)

Rust

  • Creator: Graydon Hoare, Mozilla Research
  • Why it matters: Performance without sacrificing safety

🙏 Acknowledgment of Research Community

The Mamut Lab idea would not exist without the collective work of thousands of researchers across multiple disciplines:

  • Cognitive Science: Understanding human decision-making and learning
  • Machine Learning: Neural networks, continual learning, evolutionary algorithms
  • AI Safety: Value alignment, interpretability, robustness
  • Software Engineering: Distributed systems, event-driven architectures, domain-driven design
  • Database Research: Multi-model persistence, event sourcing, CQRS
  • Programming Language Theory: Type systems, effect systems, resource management

📖 How to Learn More

All the research foundations are documented in detail:

A Note on Attribution

If we've missed crediting any research or researcher that influenced the Mamut Lab concept, please contact us at info@mamutlab.net. Proper attribution is important, and we're happy to add missing acknowledgments.

🌟 The Philosophy

Innovation is synthesis. The Mamut Lab concept doesn't claim to have invented these ideas—it aims to apply them together in a novel way to solve real problems.

The proposed value lies not in the individual components (all pioneered by brilliant researchers), but in their careful integration and practical application for long-running AI orchestration with human oversight.

Current Stage: This is an idea being prototyped, with core architectural features in early implementation. The research foundations are documented, and development is progressing transparently.

← Back to Home