Challenge 4 Explainer Session | PromptWars Virtual

Technological / product concepts & requirements (PromptWars Virtual – Challenge 4 Explainer)

What the program challenges you to build

• Bi-weekly “VIP coding” challenges: build real web applications using prompt-driven development with GenAI, with minimal heavy coding.
• Each challenge provides a problem statement; you must submit:
  1. a deployment link to a working project
  2. a LinkedIn post explaining your strategy
• GenAI usage is mandatory: the project must use GenAI in a fully functional way (not static content, no hard-coded pages, no broken/“false positive” behavior).
• Submissions should be tested thoroughly because they are evaluated both automatically and manually.

Main difference vs. a hackathon

• A cumulative leaderboard runs across the entire year—consistency matters as much as individual submissions.

Program structure / eligibility

• Open to everyone located in India
• Individual participation only (no teams)
• Three submission attempts per participant; only the latest attempt counts toward the leaderboard

---

Evaluation & scoring approach (what is being judged)

Automated code assessment dimensions

Submissions are assessed on:

• Code quality
• Security
• Efficiency
• Testing
• Accessibility
• Problem statement alignment

Manual evaluation / practical expectations

Manual review checks whether:

• GenAI is truly used
• outputs are not repetitive/incorrect
• the system has a real before/after improvement story

Guidance note: top scores aren’t just about using the “fanciest” model—they reward engineering-style solutions, including reasoning/generation rather than plain rules when appropriate.

Testing best practices emphasized in Q&A

• Testing scores can drop if edge cases aren’t covered.
• “Optimize code” guidance was used to justify improvements in efficiency/code quality (e.g., using binary search when applicable).

Deployment guidance

If GitHub Pages isn’t allowed, projects can be deployed on other platforms such as:

• Vercel
• Netlify
• Google Cloud Run
• Other hosting options were mentioned (including DigitalOcean for students)

---

Challenge 4: “Smart Stadiums and Tournament Operations” (core tutorial/analysis)

Intended application “shape”

The speaker reframes the prompt: it’s not just a fan seat-finder app. The World Cup is treated like a temporary city that must run smoothly for a few hours/days.

You must choose a persona (key design rule)

There are four personas, and the guidance is to pick one (or optionally 2) and go deep:

• Fans
• Organizers
• Volunteers (explicitly called out as often underrated)
• Venue staff

The “or” in the problem statement means you don’t need to solve everything for all four.

Optional verticals/jargon (don’t cover all at once)

The problem statement lists multiple themes; advice was to avoid implementing all:

• navigation
• crowd management
• accessibility
• transportation
• multilingual assistance
• sustainability
• operational intelligence
• real-time decision support

(“eight of them” mentioned)

Concrete examples of what each vertical could mean

• Navigation (upgraded): not just “map + chatbot”, but route planning that avoids crowd bottlenecks using live crowd data + GenAI reasoning (explain why a route is better).
• Crowd management (upgraded): GenAI should reason over CCTV/density/entry-exit data, producing explainable recommendations in plain language for ops staff (XAI-style).
• Multilingual assistance (beyond translation):
  • adapt language register/tone based on context (example: appropriate Arabic register, not just Arabic text)
• Sustainability:
  • real-time bin fill monitoring + notifications to cleaning crews (reasoning + proactive action)
• Accessibility and inclusive features:
  • step-free routing for wheelchair users
  • real-time captioning for announcements/commentary
  • mention of deaf accessibility as a differentiator
• Transportation (“lost mile”):
  • predict post-match transit demand (e.g., 40k leaving a metro station) and dynamically adjust shuttles/rides/flows
• Operational intelligence:
  • a “back office dashboard” style real-time decision support, staffing predictions, etc.

“Build it” walkthrough example (how to combine persona + verticals)

Example constructed by the speaker:

• Persona: Volunteer
• Verticals: crowd management + multilingual
• Concept: “Volunteer copilot”
  • shows live crowd density per zone/gate
  • triggers instructions when density crosses a threshold (e.g., 80% predicted)
  • redirects volunteers to gates with lower crowd
  • generates scripts in multiple languages for volunteer communication
  • includes contextual conversation handling (e.g., different guidance tone for washroom vs medical need)

Key process advice for participants

• Build for one real person moment:
  • pick persona → pick 1–2 verticals → close the loop with a demoable end-to-end flow
• Don’t build “for a year’s sake”; build to solve a specific operational pain

---

Rewards & “swag store” / credits (points system)

Currency

• Prompt Credits (PC)

How to earn credits

Earned via bi-weekly submissions and leaderboard rankings:

• Top 400 participants per challenge get prompt credits
• Credits scale with rank (higher rank → more credits)

Credits emphasize participation + consistency + documentation quality.

How credits accumulate & expire

• Credits never expire within the program year, but later messaging states:
  • Prompt credits expire on Jan 1, 2027
• Credits stack across challenges and can accumulate over time.

Redemption

Redeem PC in the swag store (items listed with approximate costs):

• caps, t-shirts, hoodies
• premium backpack
• other practical tech items (mobile stand, diary, tech organizer pouch, etc.)

Example behavior described:

• finishing high repeatedly yields enough for premium items (rank-based credit totals demonstrated)

Certificates and extra perks

• Participation certificates after 14–21 days
• Possible invitation to an in-person Prompt Wars event for top submissions

---

Q&A highlights (practical implementation guidance)

• Real-time data for crowd management
  • use public APIs and datasets; generate synthetic data if needed
  • if possible, allow evaluators to test with real uploaded datasets (CSV/PDF/SQL-like input concept)
• Tool flexibility
  • no restriction to one GenAI tool; alternatives to a specific tool were mentioned
  • credits depend on tool plans
• API keys/credits
  • they do not provide API credits or tokens
  • you choose your own free/public APIs
• Multilingual language count
  • no fixed number; include as needed, but focus on contextual relevance (based on user context such as trip location)
• Where to deploy (if not GitHub Pages)
  • Vercel/Netlify/Google Cloud Run and other hosting options recommended
• Prompt quality
  • use a prompt engineering structure (system instructions, user instructions, examples, output format)
  • optionally in JSON format
• Satisfying “GenAI matters” scoring
  • GenAI should contribute reasoning/generation (not only “fancy model” usage)

---

Main speakers / sources

• Jes: host/lead explainer of program structure, evaluation rules, and rewards overview
• Mr. Shinas: subject matter expert for Challenge 4 (Smart Stadiums & Tournament Operations), including persona/vertical breakdown and ideation
• AMA/Q&A participants were referenced (e.g., Pushba, Visha, Nihar, Kaiser, Sher Singh, Arpit Kumar, Rohit Bara), but the primary instruction sources are Jes and Mr. Shinas.