Real briefs. Real heat.
Active competitions, posted by businesses with real problems. Pick one. Ship working code. Climb the stages.
Competitions
Write a refund policy an AI agent can apply without escalating
# AI Refund Policy Agent Challenge ## Objective Create a clear, deterministic refund policy that an AI agent can autonomously apply to customer requests without requiring human escalation. ## Requirements ### Core Functionality Your refund policy must enable an AI agent to: - **Evaluate refund requests** based on objective, verifiable criteria - **Make autonomous decisions** (approve/deny) without human intervention - **Generate appropriate responses** that explain the decision to customers - **Handle edge cases** with clear fallback rules ### Policy Specifications Your policy document should include: 1. **Eligibility Criteria** - Clear conditions for refund approval (time limits, product conditions, purchase types) - Specific disqualifying factors - Required evidence or information from customers 2. **Decision Rules** - Unambiguous if-then logic - Priority ordering when multiple conditions apply - Refund amount calculation method (full, partial, prorated) 3. **Process Steps** - Information the AI must collect from the customer - Verification steps the AI can perform automatically - Actions to take after approval/denial 4. **Response Templates** - Approval message structure - Denial explanation requirements - How to maintain appropriate tone and empathy 5. **Edge Case Handling** - Ambiguous situations and their resolutions - Missing information protocols - Exception scenarios (if any) ## Success Criteria A successful policy will: - ✅ Be implementable by an AI without ambiguity - ✅ Minimize false positives (wrongful approvals) and false negatives (wrongful denials) - ✅ Require zero human escalation for cases within defined scope - ✅ Maintain customer satisfaction while protecting business interests - ✅ Be comprehensible to both AI systems and human reviewers ## Deliverable Format Submit your policy as a structured document (markdown, JSON schema, or flowchart format accepted) that an AI agent could directly parse and execute. # Success definition A support agent with no prior context can follow the policy on ten consecutive real requests and reach the same decision a competent human would, or hand the case over with a reason the customer can be told.
Build Competitive Intelligence Dashboard Prompts for Food Delivery Market Entry
Analyze Customer Churn Patterns and Build Retention Strategy
[DEV] Workspace — Pre-launch
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Spark — Flash Battle
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Workspace — Active
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Hearth — Training Arena
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Workspace — Late Grace
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Forge — Enterprise Mission
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Workspace — Evaluating
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
[DEV] Workspace — Completed
A dense fidelity-validation fixture seeded by seed-competition-states.ts. Safe to delete — prefix: [DEV].
Patient Education Materials for New Diabetes Diagnosis
Valley Health Clinic serves a diverse patient population and has identified that newly diagnosed Type 2 diabetes patients often leave their first endocrinology appointment overwhelmed and unclear on next steps. The clinic needs a comprehensive but accessible patient education packet that covers the first 90 days post-diagnosis. Your task is to create educational materials that explain: what Type 2 diabetes means in plain language, blood glucose monitoring basics, medication adherence, dietary principles (not a meal plan), physical activity recommendations, warning signs requiring immediate care, and how to prepare for follow-up appointments. The materials should be appropriate for a general adult audience with varying health literacy levels (target 8th grade reading level), culturally neutral, and organized in a logical learning sequence. The clinic will print these materials in-house, so the format should be straightforward to produce. Content must empower patients without causing undue alarm.