prompt
FreeStructured divergent ideation for open-ended, high-stakes problems
FreeFree tier
Inputs: textOutputs: text
About prompt
The Divergent Ideation Architect is a structured system prompt designed for large language models to perform high-quality divergent thinking. It implements a two-phase loop: Diverge (isolated parallel generation of multiple ideas) followed by Focus (scoring, clustering, trap detection, and deepening). A pre-flight gate ensures the technique is only applied to truly open-ended, high-stakes problems. Based on the ADHD framework (from UditAkhourii/adhd), it has been shown to improve idea breadth, novelty, and trap detection over single-shot baselines. The prompt is hosted in the ai-boost/awesome-prompts GitHub repository and is freely available as open source.
Key Features
Two-phase divergent ideation loop: Diverge (parallel generation) then Focus (scoring, clustering, trap detection)
Pre-flight gate to ensure technique is only used when appropriate (open-ended, high-stakes problems)
Mechanically separated generator and critic roles to avoid strangling idea quality
Evaluated across 6 open-ended engineering problems with improvements in breadth (+4.17), novelty (+5.17), and trap detection (+7.67) over single-shot baseline
Explicit invocation mode via '/diverge' command for opt-in use
Pros & Cons
Pros
- Significantly improves idea quality (breadth, novelty, and trap detection) compared to standard single-shot prompting
- Structured process reduces premature convergence in autoregressive reasoning
- Includes a self-judging pre-flight gate to avoid unnecessary cost
- Free and open source, available on GitHub
Cons
- Expensive to run: approximately 10 agent calls, 30–90 seconds wall clock time, 5–10× the cost of a single answer
- Not suitable for simple or canonical questions; requires open-ended, high-stakes problems to be worthwhile
- Requires a capable LLM (e.g., GPT-4 class) with support for multi-step prompting and role separation
Best For
Open-ended design decisionsFuzzy debuggingNaming thingsAPI surface designStrategy explorationAny prompt of the form 'give me a few ways to…'
FAQ
What is the Divergent Ideation Architect prompt?
It is a system prompt for AI assistants that guides them through a structured two-phase divergent ideation process: first generating many ideas in parallel (diverge), then evaluating, clustering, and deepening them (focus). It is designed to replace textbook answers with deeper exploration of the design space.
When should I use this prompt?
Use it when you have an open-ended, high-stakes problem where the cost of a single obvious answer being wrong is high. Examples include architecture decisions, naming, API design, or strategy. The prompt includes a pre-flight gate that will abort if the question is not suitable.
How much does it cost to run?
The prompt itself is free and open source. However, executing it requires multiple API calls to an LLM (about 10 calls, 30-90 seconds), making it 5-10 times more expensive than a single answer. Use it only for problems that genuinely benefit from divergent thinking.
Where can I find the prompt?
The prompt is hosted in the ai-boost/awesome-prompts GitHub repository at https://github.com/ai-boost/awesome-prompts/blob/main/prompts/divergent_ideation_architect.txt . It is freely downloadable and modifiable.