The Vision
What if you could see what's happening around the world in real time? Not a feed. Not a newsletter. A live, interactive map that shows where major events are unfolding geographically. AI breakthroughs, economic shifts, geopolitical conflicts, public health crises. A real-time command center for global awareness.
That was the idea behind PulseView. The harder question wasn't what to build. It was whether it could be built in 24 hours.
The answer turned out to be yes.
The Toolchain
Two AI agents worked in concert, each playing a different role:
Perplexity Computer (Hours 0-6)
The architect, researcher, designer, and first coder. Through natural conversation, it designed the full-stack architecture, generated logos, researched domain names, wrote the Convex backend, Cloudflare Worker frontend, and the entire interactive map UI.
OpenAI Codex (Hours 6-24)
The tireless implementer. After the foundation was committed, Codex handled rapid iteration: refactoring, adding features, fixing bugs, and scaling the system from prototype to production. 16 commits over 18 hours.
Key Technologies
- Convex — Real-time database, cron jobs, and serverless functions
- Cloudflare Workers — Edge hosting with caching and API protection
- Perplexity Sonar API — AI-powered news intelligence for hourly content
- D3.js + TopoJSON — Interactive map visualization with diamond markers
The Build
Phase 1: Foundation
The first six hours covered concept, naming, branding, frontend, and backend architecture. Perplexity Computer brainstormed the concept, researched 60+ domain names, generated logo iterations, established the color palette (steel blue + gold), and built the interactive world map with D3.js.
By hour six, the full Convex backend was in place: event schema, hub promotion logic, Perplexity Sonar integration with URL deduplication, hourly cron job, REST API endpoints, and 40 historical seed events. All committed to GitHub in a single push.
Phase 2: Production Polish
OpenAI Codex took over and delivered 16 commits that transformed the prototype into a production app:
- Ingest pipeline overhaul — 4,400+ lines changed. Rewrote the news fetching pipeline for production reliability with comprehensive logging and a city hero image system
- Multi-topic system — Refactored from a single AI topic to a fully database-driven multi-topic architecture. Expanded to 7 topics: AI News, Global Economy, Global Security, US Politics, US Economy, US Healthcare, and Top Stories
- Top Stories curation — 585+ lines of a new AI-powered story ranking system that aggregates and surfaces the most important stories
- Mobile experience — Hybrid feed as the default mobile view with responsive layout
- Infrastructure hardening — Cloudflare Worker edge proxy, per-IP rate limiting, edge cache invalidation, and optimized asset caching
The Final Product
PulseView is live and deployed. Here's what it does:
- Interactive global map showing worldwide news activity in real time
- Diamond markers with category-colored indicators for each event
- Heatmap coloring showing hub activity density (gold = hot)
- 7 topics spanning AI, economics, security, politics, health, and curated top stories
- Live mode with hourly auto-refresh and historical mode with date range picker
- Article modals with hero images, summaries, and source links
- Light and dark mode with system preference detection
- Production-grade infrastructure with edge caching and API protection
By the Numbers
| Metric | Value |
|---|---|
| Total time | 24 hours |
| Commits | 18 |
| Topics | 7 |
| Convex backend modules | 14 |
| Frontend code | 2,500+ lines |
| News ingestion logic | 900+ lines |
| Top stories curation | 500+ lines |
| AI agents | 2 |
| Humans | 1 |
| Lines of code written by hand | 0 |
What I Learned
PulseView is an example of how quickly you can turn an idea into a product. The traditional development cycle of weeks or months compressed into a single day, not by cutting corners, but by leveraging AI agents that can think architecturally, write production code, and iterate at machine speed.
Perplexity Computer handled the creative and architectural work: naming, branding, system design, and the initial implementation. OpenAI Codex took over for the grinding iteration: refactoring, scaling, hardening, and polishing. Together, they covered the full spectrum of software engineering.
What's remarkable isn't just the speed. It's the quality. PulseView has edge caching, multi-topic support, AI-powered content curation, mobile responsiveness, and production infrastructure. These aren't shortcuts. They're features that typically take teams weeks to implement.
The human brings the taste, the judgment, the "make it diamonds not circles" moments. The AI brings the tireless execution.
The full case study with detailed build timeline, architecture diagrams, and the complete breakdown is available as a PDF below.
Download the Full Case Study
PulseView: From Idea to Product in 24 Hours — 10-page PDF with the complete build timeline, architecture overview, and detailed breakdown of every phase.