Overview
ADHD is a groundbreaking skill for coding agents, designed to overcome the limitations of linear reasoning. It employs a novel Tree-of-Thought approach with advanced pruning techniques, built upon the Claude and Codex Agent SDK. By fanning out parallel, divergent thoughts across distinct cognitive frames, ADHD enables agents to explore a wider solution space, score and prune suboptimal paths, and cultivate the most promising ideas. This makes it an indispensable tool for creative problem-solving and interdisciplinary work.
Key Features
- Parallel Divergent Ideation: Generates multiple, independent reasoning threads under different cognitive frames, avoiding premature convergence and anchoring.
- Advanced Pruning & Scoring: Utilizes a critic pass to rigorously score, cluster, and prune suboptimal ideas and identify potential "traps" before they waste resources.
- Deepened Survivors: Focuses on developing the most promising insights, providing more robust and innovative solutions.
- Architectural Fix for Reasoning: Tackles the problem of anchoring and premature convergence at an architectural level, rather than relying solely on prompting techniques.
- Cognitive Frames: Leverages a diverse set of deliberately distorted cognitive frames to encourage truly novel perspectives.
Typical Use Cases
ADHD excels in scenarios demanding creativity, breadth of exploration, and robust decision-making, including:
- Design Decisions: Exploring diverse architectural options, API surface designs, and strategic approaches.
- Fuzzy Debugging: Uncovering root causes and potential solutions for complex, ill-defined problems.
- Naming and Branding: Generating a wide array of creative and fitting names.
- Strategy Development: Brainstorming and evaluating multiple strategic paths.
- Ideation and Brainstorming: Answering prompts of the shape "give me a few ways to..." with exceptional depth and novelty.
