Permanent weekly record
Weekly AI Futures Calibration: October 12, 2026
This page preserves the distribution published on October 12, 2026 and the primary evidence that justified its revision. It is not rewritten with later information.
Published distribution
Level 1: Existential Loss of Control
- Probability
- 12.9%
- Prior calibration
- 13.1%
- Change
- -0.2 percentage points
Level 2: Concentrated Dystopia / Techno-Feudalism
- Probability
- 24.8%
- Prior calibration
- 24.8%
- Change
- 0 percentage points
Level 3: Turbulent Transition
- Probability
- 35%
- Prior calibration
- 35.4%
- Change
- -0.4 percentage points
Level 4: Managed Partnership
- Probability
- 16.7%
- Prior calibration
- 16.5%
- Change
- +0.2 percentage points
Level 5: Broad Abundance & High Agency
- Probability
- 10.6%
- Prior calibration
- 10.2%
- Change
- +0.4 percentage points
| Horizon | Probability | Prior calibration | Change |
|---|---|---|---|
| Level 1: Existential Loss of Control | 12.9% | 13.1% | -0.2 percentage points |
| Level 2: Concentrated Dystopia / Techno-Feudalism | 24.8% | 24.8% | 0 percentage points |
| Level 3: Turbulent Transition | 35% | 35.4% | -0.4 percentage points |
| Level 4: Managed Partnership | 16.7% | 16.5% | +0.2 percentage points |
| Level 5: Broad Abundance & High Agency | 10.6% | 10.2% | +0.4 percentage points |
Accepted primary evidence
These signals cleared the date, provenance, novelty, and materiality filters. Their summaries preserve the caveats in the public record.
Anthropic validates high-severity vulnerability findings from an open-source scanner
Anthropic reports more than 29,000 candidate vulnerabilities from six months of scanning. In a validation sample, expert penetration testers reviewed 97 high- or critical-severity findings across 48 projects: 85 met the coordinated-disclosure bar, 11 were real but duplicated known issues or other scan findings, and one was invalid. The result shows useful defensive capacity and a growing human-review bottleneck, but it remains a company-run evaluation rather than an independent population estimate.
Declared horizon impact
- +1 Existential Loss of Control
High-quality automated vulnerability discovery is dual use and can expand cyber capability faster than review capacity
- +1 Turbulent Transition
Human validation and disclosure capacity is becoming a bottleneck as automated discovery scales
- +2 Managed Partnership
Expert triage, coordinated disclosure, and maintainer remediation provide a concrete human oversight pathway
Primary sources
Applied Digital reports 250 MW live and 1.41 GW under long-term data-center leases
Applied Digital's October 8 filing reports that 250 MW of critical IT load was live at Polaris Forge 1 after an October 1 delivery. The company had leases for about 1.41 GW across five campuses, representing about $36 billion in contracted base-term revenue, and recorded $65.8 million in quarterly base rent. These are delivered capacity, signed leases, and reported revenue, not the separate optional 1 GW Finland power agreement. The filing also reports $6.4 billion in debt.
Declared horizon impact
- +3 Concentrated Dystopia / Techno-Feudalism
Delivered capacity and long-term lease revenue remain concentrated in capital-intensive campuses and hyperscaler relationships
- +1 Turbulent Transition
Rapid construction, grid demand, and debt-financed expansion add physical and financial transition pressure
Primary sources
OpenAI publishes a large catalogue of model-generated mathematical research
OpenAI released a public catalogue containing 722 manuscripts in 372 families. The repository says the model was given about 4,000 problems and used an average of three hours of ChatGPT Pro thinking compute per result. The catalogue includes different stages of verification, and OpenAI warns that some results without Lean formalization may contain errors. The release therefore supports scientific acceleration, but not a claim that every manuscript is independently verified.
Declared horizon impact
- +1 Turbulent Transition
A large research output increases pressure on peer review, attribution, and scientific verification systems
- +1 Managed Partnership
Public manuscripts, revision protocols, citations, and partial formalization create a checkable governance path
- +3 Broad Abundance & High Agency
Broad model-generated mathematical research raises the prospect of widely useful scientific acceleration