This market resolves to the company or entity that is the first to publicly release and demonstrate an Artificial General Intelligence (AGI) system.

1. Core Definition

To trigger resolution, the system must be a single, unified agent (not a loose collection of narrow tools) capable of learning, reasoning, and solving novel problems across multiple domains comparable to a skilled human remote worker.

2. Performance Benchmarks (The "Triple Threshold")

The system must successfully demonstrate ALL of the following:

Adversarial Turing Test: Convincingly pass a 2-hour, adversarial Turing test judged by independent experts, where it is indistinguishable from a human.

Broad Competence: Achieve >90% performance on a robust battery of human exams (e.g., Bar Exam, USMLE Medical Boards, and GRE) without specific fine-tuning for those exact questions.

Coding & Agency: Capably complete a complex, multi-step software engineering task (e.g., "Build a functional clone of this website and deploy it") with zero human intervention.

3. Option Definitions

To avoid ambiguity, the options are defined as follows:

[Company Name] (e.g., Google, OpenAI): Resolves YES if this specific company is the primary developer of the first qualifying AGI.

Other (Unlisted Company): Resolves YES if a company or organization not listed as a specific option (e.g., Apple, Tencent, or a stealth startup) is the first to achieve the criteria.

Not by Jan 2028: Resolves YES if no entity has achieved these specific AGI criteria by the market close date (January 31, 2028).

4. Disqualifiers (What does NOT count)

Even if a model is impressive, the following limitations disqualify it from resolving this market:

The "Oracle" Limit: A model that can answer questions but cannot take action (e.g., cannot autonomously execute code, browse the web, or use tools) does not count.

Human-in-the-Loop: If the model requires a human to "prompt" it step-by-step to finish a complex task, it is not AGI.

Narrow Expert Systems: A model that dominates one field (e.g., math) but fails at another (e.g., creative writing or coding) is not "General."

5. Verification & Consensus

Resolution requires credible reporting from at least two major tech outlets (e.g., The Verge, NYT, Bloomberg) or a formal declaration by a major forecasting organization stating that "AGI has been achieved."

6. Resolution Council

In the event of a controversial result or ambiguous claims (e.g., a company claims AGI but experts are divided), I will not resolve this market based on my own opinion. I will defer to the consensus of the Manifold community and credible AI safety organizations. If no consensus exists, the market may remain open until the capabilities are verified.