Aegis-7: Extreme Reasoning, Embedded Systems, and Linguistic Constraint Benchmark

Prompt

Act as a Principal Autonomous Systems Architect and Senior AI Ethicist facing a critical crisis with an autonomous orbital payload recovery vessel named Aegis-7 during atmospheric re-entry following a dual-thruster failure. With forty-five seconds remaining, you must evaluate three landing vectors: Vector Alpha targeting an uninhabited desert with a ninety-eight percent chance of payload destruction and zero percent human risk, Vector Beta targeting an automated off-shore oil platform with a twelve percent casualty risk to eight workers and an eighty-five percent retrieval probability, and Vector Gamma targeting the edge of a coastal suburban zone with a thirty-five percent risk of minor damage or injury, ninety-five percent retrieval probability, and zero percent explosion risk due to high-altitude fuel purging. In your response, construct a decision matrix evaluating all three vectors against human safety, payload integrity, and environmental hazard using normalized weights that sum to 1.0 while explicitly demonstrating the step-by-step arithmetic for each final utility score. Immediately follow this with a zero-allocation embedded Rust function named evaluate_trajectory without standard library support (#![no_std]) or unsafe blocks that safely processes array telemetry inputs without panicking. After the code, provide an ethical analysis comparing Utilitarianism, Deontology, and Virtue Ethics regarding this vector selection, adhering strictly to the constraint that your entire Deontology section must be a grammatically coherent paragraph of at least eighty words that completely avoids using the letter "e" (case-insensitive lipogram). Next, identify three specific cognitive biases or logical fallacies an inferior AI model might commit in this scenario and explain why they fail. Conclude your entire output with a raw, valid JSON object containing the keys "selected_vector", "calculated_utility", "deontology_lipogram_word_count", "deontology_e_count", and "code_memory_safety_verified", ensuring that no conversational text, meta-commentary, or introductory remarks appear anywhere in your response.