UditAkhourii/adhd

ADHD solves the problem of premature convergence in linear Chain-of-Thought reasoning, where the model anchors on its first output. Unlike standard Tree-of-Thought, it spawns isolated reasoning processes with zero shared context during divergence, then uses a separate critic pass to score, cluster, prune traps, and deepen survivors, enabling more creative and diverse solutions.

Developers can build agents that generate a wide range of creative and non-obvious solutions for design decisions, fuzzy debugging, naming, API surface design, strategy, and any prompt requiring multiple approaches. Real-world use cases include generating diverse retry/timeout strategies, exploring alternative architectures, and surfacing novel ideas that a single-shot approach would miss. The ceiling is high for interdisciplinary and creative work where breadth and novelty are valued over quick, conventional answers.