# Pulpie: Web Content Extraction API — 90-Second Demo Script

## [0–10 sec] Problem Hook  
"Extracting clean content from HTML. Current tools cost $159,000 to process a billion pages. We do the same job for $7,900. Same quality. One-twentieth the price."

## [10–30 sec] The Mechanism  
"Pulpie is encoder-based. One forward pass over the HTML. We label each block: boilerplate or content. Return clean markdown. No token-by-token decoding. No latency spiral. Just fast, efficient extraction."

## [30–50 sec] The Math  
"$7,900 for a billion pages with Pulpie. $159,000 with the leading alternative. We matched their extraction quality in benchmarks. Then undercut the cost by 20x. Download the benchmark report; see it yourself."

## [50–70 sec] Where It Works  
"RAG training data pipelines. Bulk web scraping. Data enrichment workflows. Anywhere extraction happens at scale. News archives. Research papers. Tech blogs. Every page returns signal, no noise."

## [70–90 sec] Next Step  
"Try Pulpie on your extraction job. Check the quality. Look at the invoice. Then switch. Free tier, no credit card. Full API docs at pulpie.dev."

---

## Director's Notes
- **Tone**: Numbers-first, confident, no hype
- **Pacing**: Slow enough to read; fast enough to hold attention
- **Visuals to capture**:
  - [Sec 0–10] Split-screen: messy HTML → clean markdown
  - [Sec 10–30] Animation: encoder (single forward pass) vs. decoder (token-at-a-time loop)
  - [Sec 30–50] Cost table: Pulpie $7.9k, competitor $159k, same quality
  - [Sec 50–70] Real extracted pages: before/after from news, research, SaaS docs
  - [Sec 70–90] Pricing page with free-trial button and no-card promise
- **CTA**: "Start free. No card. pulpie.dev/try"
- **Duration**: ~45 seconds actual talk, ~90 seconds with visuals and pauses