Paid Trial — Use Codex, Claude, or Cursor to Build a Mini App with MinuteWork CLI — No Manual Coding

## Overview We’re hiring AI-native developers for a short, paid evaluation of the **MinuteWork CLI and Builder workflow**. You will choose an AI-native business idea from our catalog. MinuteWork’s Builder guidance will then help **Codex, Claude Code, or Cursor** transform your idea into a bounded autonomous sprint and execute it. You will **not** write the sprint or application code yourself. Your role is the **founder and AI-agent operator**: - Choose a business concept - Make a few product decisions - Approve the AI-generated sprint - Let the coding agent execute it autonomously - Handle human-only login and approval steps - Verify the final evidence - Report what impressed, confused, or blocked you We are evaluating the complete journey: **Business idea → autonomous sprint → generated workspace → working prototype or verified blocker** --- # Choose a Business Idea Select one of the following ideas. ## FoodRun Office Lunch Coordinator Collect lunch requests, group orders by restaurant, enforce a cutoff, prepare a grouped-order draft, and assign a volunteer runner. ## PropertyCare Maintenance Concierge Receive maintenance requests, classify urgency, gather evidence, and prepare remediation tasks. ## VendorReady Compliance Desk Collect vendor documents, identify missing or expiring items, and create review tasks. ## ClientLaunch Agency Onboarding Collect client goals and assets, track missing inputs, and prepare an onboarding plan. ## FleetFix Maintenance Coordinator Receive vehicle issue reports, classify urgency, and prepare service-request drafts. ## TurnoverOS Rental Coordinator Convert inspection findings into cleaning and repair tasks and prepare an approval packet. ## ProposalDesk Turn an opportunity brief into research tasks, an evidence checklist, and a proposal draft for review. ## ContentOps Review Studio Turn a campaign request into content drafts, review tasks, revisions, and a publication-ready package. These are synthetic prototypes. They do not involve: - Real payments - External dispatch - Production publication - Customer credentials - Autonomous spending --- # How the Evaluation Works 1. Initialize a clean MinuteWork workspace. 2. Open it in Codex, Claude Code, or Cursor. 3. Invoke the MinuteWork Autonomous Sprint Builder. 4. Select and personalize one idea by choosing: - The target customer - The business name - One distinguishing feature 5. Answer up to five short business questions. 6. Let the agent generate a verification-first autonomous sprint. 7. Review only the sprint’s: - Business intent - Scope - Safety boundaries 8. Approve the sprint or ask the agent to regenerate it if it misunderstood the business. 9. Give the agent the following instruction: Execute the approved sprint continuously. Make reasonable in-scope technical decisions and verify your own work. Do not wait for me unless authentication, credentials, spending, publication, or a material product decision requires a human. 10. Let the agent: - Build - Validate - Compile - Test - Diagnose failures - Attempt the permitted preview path 11. Stop after two hours or when the agent proves a terminal platform blocker. The two-hour window includes: - Sprint generation - Implementation - Verification - Reporting --- # What the AI-Generated Sprint Contains MinuteWork—not you—will generate: - The intended business outcome - A bounded MVP - Explicit non-goals - User and operator journeys - Records, actions, workflows, and views - A dependency-aware work sequence - Acceptance tests - Evidence required before claiming completion - Human-only decisions - Security and privacy boundaries - Stop conditions - Final verification and handoff instructions --- # Absolute Rule: No Manual Coding or Planning ## You May - Choose and personalize an idea - Answer product questions - Correct a misunderstanding of the business - Approve or regenerate the sprint - Complete authentication - Approve safe agent actions - Make human-only product or safety decisions - Tell the agent to continue, diagnose, or verify - Stop an unsafe or out-of-scope action ## You May Not - Write or paste application code - Manually edit source or configuration files - Manually write or repair the technical sprint - Debug or repair the implementation yourself - Patch dependencies or `node_modules` - Provide code snippets from another source - Replace failed MinuteWork paths with custom infrastructure - Hide failures - Manufacture success using mocks Every technical planning artifact and source-code change must be generated by **Codex, Claude Code, or Cursor**. --- # Deliverables Submit the following: - Your selected idea and customizations - The initial Builder prompt - The complete AI-generated sprint - The complete agent transcript - The generated repository, even if incomplete - A 5–10 minute screen recording The recording should show: - Idea selection - Sprint generation - Autonomous execution - The final result or blocker --- # Evaluation Report Submit a short evaluation answering the following questions: 1. Did the sprint understand your business? 2. What did you regenerate or correct? 3. How often did the agent need you? 4. What did it accomplish autonomously? 5. Where did it become confused? 6. What was the most impressive moment? 7. What were the three greatest points of friction? 8. Would you independently consider MinuteWork for a future client project? Why or why not? There is no required positive answer. --- # Privacy and Evidence Requirements Remove the following from all submissions: - Tokens - Passwords - `.env` contents - Personal data - Private filesystem paths Do not include confidential credentials or sensitive information in recordings, repositories, transcripts, or reports. --- # Payment - **Fixed price:** $100 - **Milestone:** One fully funded milestone - **Maximum evaluation window:** Two hours - **Unpaid overrun:** None A verified blocker qualifies as a completed evaluation. Payment is for following the autonomous process and delivering the required evidence—not for forcing the prototype to succeed. Strong participants may be invited to additional paid evaluations. --- # Who Should Apply You should: - Regularly use Codex, Claude Code, or Cursor - Understand Git, terminals, Node.js, and modern web applications - Be comfortable letting an agent make routine technical decisions - Know how to verify an agent’s completion claims - Be disciplined enough not to take over manually - Produce clear, reproducible technical feedback - Already have lawful access to your chosen coding agent Please answer: 1. Which coding agent will you use? 2. Which two business ideas interest you most? 3. What have you built primarily by directing an AI coding agent?