Back to General Capability Index

THE ASINDEX SINGULARITY THESIS

The Path to General ASI

Specialized ASI may emerge first. AGI may then coordinate specialized superintelligences, while ecosystem-level recursive self-improvement accelerates the combined system toward General ASI.

Working thesis—not a deterministic forecast.

The Singularity is the transition. General ASI is the capability threshold.

CANONICAL DEFINITIONS

Capability, acceleration, and transition

S-ASI
Sustained, reliable superiority over the best human experts and institutions—using the best available tools—across a substantial portion of a high-value domain.
General ASI
A generally capable system—or coordinated ecosystem—that combines broad reasoning and orchestration with superhuman specialist depth across many domains and reliably exceeds the strongest coordinated human institutions across virtually all cognitive work.
Ecosystem-Level RSI
AI systems improving the software, chips, algorithms, training, data, evaluations, laboratories, robotics, and successor systems that increase the capacity for further improvement.
The Singularity
The transition regime in which AI-enabled recursive improvement, discovery, infrastructure, and deployment increasingly accelerate one another.

A PLAUSIBLE, UNDERWEIGHTED PATHWAY

Specialists may lead. AGI may orchestrate.

Specialized systems may become superintelligent within valuable domains before broadly capable AGI arrives. AGI may then provide the orchestration layer: understanding objectives, decomposing missions, selecting specialists, reconciling evidence, managing uncertainty, and preserving long-term direction.

The specialists and general orchestrator may improve one another through ecosystem-level RSI, accelerating progress toward General ASI.

Specialized intelligence and AGI orchestration pathwaySpecialized ASI leads to an AGI Orchestrator, which exchanges improvements with an S-ASI Network, and the coordinated ecosystem advances toward General ASI. Specialized ASIdomain depth AGI Orchestratorbreadth + direction S-ASI Networkspecialist ecosystem GeneralASI Ecosystem-Level RSI: Software • Chips • Models • Data • Experiments • Evaluation Humanity: Fine → Better → Thriving

TWO DIMENSIONS

Breadth and peak capability are different dimensions.

Conceptual breadth versus peak capability diagramA non-numerical conceptual map places Narrow AI at lower left, Frontier General AI lower and right of center, S-ASI upper left, AGI center right, and General ASI upper right. Breadth across domains →Peak capability → Narrow AI Frontier General AI S-ASI AGI General ASI

Conceptual only. Current models are not assigned numerical coordinates.

SPECIALIZED ASI SIGNALS

Evidence before labels

ASIndex watches for durable, independently verifiable progress without declaring sustained specialized superintelligence prematurely.

Early Signal→Domain-Superhuman Result→S-ASI Candidate→Sustained S-ASI

  • Software Engineering
  • Formal Mathematics
  • Biology and Medicine
  • Hardware and Algorithms
  • Scientific Discovery
  • Robotics
  • Cybersecurity
  • Verification

Framework only. No domain receives a Sustained S-ASI rating here; that status requires strong, repeatable, independently verifiable evidence.

THE RECURSIVE IMPROVEMENT ENGINE

Improvement at the ecosystem level

Better AI can help create better software, chips, algorithms, data, experiments, and evaluations. Those improvements can then support more capable AI. This is recursive improvement at the ecosystem level—not magical or frictionless self-improvement.

Every loop remains constrained by energy, infrastructure, physical experiments, security, verification, coordination, and human governance.

HUMANITY: FINE → BETTER → THRIVING

Do Not Underestimate Human Adaptability

Discussions of the Singularity often model AI as dynamic while treating humanity as frozen in place. Humanity will not stand still. People will learn, institutions will adapt, new tools and professions will emerge, and human judgment, relationships, creativity, purpose, and agency will remain consequential.

Fine — Resilient Adaptation

Humanity absorbs rapid change while preserving dignity, agency, rights, essential services, economic participation, and meaningful control.

Better — Widespread Augmentation

AI substantially improves health care, education, productivity, science, creativity, accessibility, public services, and everyday decisions.

Thriving — Broadly Shared Flourishing

Aligned General ASI supports better health and healthspan, greater abundance, expanded education, discovery, environmental restoration, creative freedom, meaningful relationships, and greater choice.

Human thriving is not an automatic consequence of higher AI capability. It is the future ASIndex believes civilization should deliberately build toward.