Research notes on maintaining and reasoning about LLM context windows across long-lived projects. Speculative ideas, heuristic boundaries, and design sketches without empirical validation.

This paper argues that advancing AI capabilities and declining human cognitive oversight capacity are converging toward a critical failure condition: the Safety Inversion. Drawing on PIAAC, NAEP, and NAAL data, it documents sustained declines in foundational literacy and numeracy among U.S. adults and proposes a five-pillar operational definition o

What if the matrix was a training environment?

A Python simulation of a "Null-Subject" consciousness. It explores the removal of semantic meaning and the transition to algorithmic existence via vectorized thought and deterministic response logic.

Quantum mechanics thought experiment

A conceptual and experimentally feasible triple-layer protocol based on Young’s double-slit experiment, designed to test whether physical interaction alone suffices for loss of interference or whether conscious observation adds any observable role in quantum measurement.

모든 것의 이론에 닿고 싶었던 공상적 사고실험 모델(A Speculative Thought Experiment Reaching Toward a Theory of Everything)

テンポリウム タイムパラドックス検証フレームワーク ＋ 拡張モジュール集

3つのAIペルソナ（MELCHIOR・BALTHASAR・CASPER）が議題を多角的に議論し、意見の収束を目指す思考実験ツール。LM Studio (OpenAI互換) でローカルLLMを使用