stindardlogic/brainstorming-ideation-sft-100k

Brainstorming and Ideation SFT (100K)

100,000 ShareGPT conversations demonstrating structured, high-quality brainstorming and ideation across 22 professional domains. Each example takes a realistic context and constraint, then generates specific, actionable, well-reasoned ideas — not generic advice dressed as creativity.

Motivation

Brainstorming and ideation is one of the highest-value use cases for AI assistants, and one where models routinely underperform:

• Generic ideas: "Create content" instead of "Launch a CFO benchmark report targeting finance directors who search for expense policy templates"
• No constraint reasoning: Ideas that ignore the stated constraints (budget, team size, timeline)
• Quantity over quality: 20 mediocre ideas instead of 10 specific, actionable ones
• No prioritization signal: Treating a high-impact idea and a marginal one with identical weight
• Missing the "why": Stating what to do without explaining why it works in this specific context
• Ignoring existing assets: Not leveraging what the person already has (email list, community, team skills)

This dataset trains models to produce ideas that are concrete, constraint-aware, and grounded in the specific situation — the kind of ideas a smart colleague with domain expertise would generate.

Dataset Description

100,000 conversations across 22 professional domains:

Domain	Examples
product_features	AI-native features for a project management tool
marketing_campaigns	Organic content strategy for DTC supplement brand
startup_pivots	Adjacent markets for a legal AI startup
team_building	Remote team cohesion without mandatory fun
revenue_growth	Revenue strategies without increasing ad spend
content_strategy	B2B thought leadership with small content team
app_features	DAU/MAU improvement for a personal finance app
hiring_strategy	Competing with FAANG without matching equity
problem_solving	Hospital no-show rate reduction
workshop_facilitation	Strategy offsites that produce concrete outputs
personal_productivity	Deep work reclamation for an engineering manager
go_to_market	Devtools GTM strategy for self-serve model
cost_reduction	20% opex reduction without layoffs
educational_program	CS curriculum for the AI era
community_building	Converting passive OSS community to active contributors
side_project	Backend engineer side project ideas with revenue models
research_directions	PhD dissertation ideas at AI × cognition intersection
event_planning	Founder summit experiences that create real connections
crisis_communication	Data breach communication strategy
nonprofit_fundraising	Growth strategy for urban tree planting nonprofit
ux_improvements	Churn reduction for a B2B analytics dashboard
business_models	OSS monetization without alienating the community

Plus: investor_pitch, mental_models in extended scenarios.

Format

{
  "conversations": [
    {
      "from": "human",
      "value": "I need help brainstorming ideas. Here's the context:\n\n**Situation:** A B2B SaaS company...\n\n**Constraints:** Focus on AI-native features...\n\n**Request:** Generate 10 innovative AI-native feature ideas..."
    },
    {
      "from": "gpt",
      "value": "Here are 10 ideas for your consideration:\n\n1. **Meeting-to-Task Autopilot** — Joins Zoom/Meet calls, listens to discussions, auto-creates tasks...\n\n2. **Dependency Risk Forecasting** — Analyzes historical project data to predict which task dependencies..."
    }
  ],
  "metadata": {
    "domain": "product_features",
    "format_note": "numbered list with bold title and description for each idea",
    "idea_count": 10
  },
  "id": "abc123"
}

Key Properties of Responses

1. Constraint-aware ideas: Every idea is checked against the stated constraints (budget, team size, timeline, resources). No ideas that require a sales team when the company explicitly has none.

2. Specific over generic: Ideas include concrete execution details — actual price points, specific platforms, realistic metrics, named tools and channels. Not "use social media" but "post Show HN on Hacker News the day of launch."

3. Domain expertise embedded: Ideas reflect real practitioner knowledge — the fundraising ideas reference actual grant programs (USDA Urban and Community Forestry); the devtools GTM ideas reference actual channels (Product Hunt, HN, GitHub Marketplace).

4. Implicit prioritization: The most actionable, highest-impact ideas are structured first. Ideas build from quick wins to larger strategic bets.

5. Rationale embedded: Each idea explains why it works, not just what to do. The "why" is domain-specific — not a generic "this will increase engagement" but "loss aversion drives daily opens to check status."

6. Leverages existing assets: Ideas reference and build on what the person already has — the email list, the community, the team skills, the existing customers.

Use Cases

• SFT fine-tuning for AI assistants focused on strategic thinking and business problem-solving
• Training AI brainstorming tools for product managers, founders, and business operators
• Building AI strategy advisors for consulting and advisory workflows
• Improving model performance on creative and strategic reasoning tasks
• Training models for executive coaching and leadership development platforms
• Fine-tuning models for AI-powered workshopping and decision support tools

License

Apache 2.0