stemedb/cmd/pitch-voiceover/script.go
jordan 157dbbb9eb feat: Complete Aphoria Phase 8-9 + UAT suite (90/90 tests passing)
## Phase 8: Enterprise Extractor Improvements 
- 14 security extractors (TLS, JWT, SQL injection, XSS, etc.)
- 10 framework-specific extractors (Spring, Django, Rails, etc.)
- Config file security detection (YAML, TOML)

## Phase 9: Autonomous Extractor Generation 
- Shadow mode executor with TP/FP tracking
- Graduation pipeline with confidence thresholds
- Auto-rollback on regression detection
- Cross-project pattern syncing

## UAT Suite Complete (14 scripts, 90 tests)
- test-core-detection.sh (6 tests)
- test-declarative-extractors.sh (5 tests)
- test-domain-frameworks.sh (5 tests)
- test-domain-unreal.sh (3 tests)
- test-llm-extraction.sh (6 tests)
- test-eval-harness.sh (5 tests)
- test-cross-language.sh (3 tests)
- test-precommit-performance.sh (4 tests)
- test-output-formats.sh (8 tests)
- test-drift-detection.sh (6 tests)
- test-exit-codes.sh (12 tests)
+ 3 more scripts

## Other Changes
- Updated roadmap to mark Phase 8-9 complete
- Added .gitignore entries for build artifacts
- Updated pre-commit: 800 line limit, exclude tests/data/cmd

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-06 22:50:55 -07:00

218 lines
7.2 KiB
Go

package main
// SpeakingBlock represents a single voiceover segment.
type SpeakingBlock struct {
Number int
Slug string
Step string
Text string
}
// Script contains all speaking blocks for the pitch demo.
// Rewritten for natural, conversational delivery.
var Script = []SpeakingBlock{
// HOOK - Lead with the pain
{
Number: 1,
Slug: "slide1-hook",
Step: "Slide 1 Hook",
Text: "In 2024, seventy-nine percent of FDA Warning Letters cited data integrity failures. The core issue? Companies couldn't reconstruct who did what, when, or why. The audit trail was missing.",
},
{
Number: 2,
Slug: "slide1-reveal",
Step: "Slide 1 Reveal",
Text: "And now, with AI making more of the decisions, that audit trail matters more than ever.",
},
// PROBLEM - Why this keeps happening
{
Number: 3,
Slug: "slide2a",
Step: "Slide 2a",
Text: "Here's the thing about your data warehouse. It stores the current answer. But what happens when your sources don't actually agree?",
},
{
Number: 4,
Slug: "slide2b",
Step: "Slide 2b",
Text: "When a study gets retracted, which decisions did it affect? Can you answer that question today?",
},
{
Number: 5,
Slug: "slide2c",
Step: "Slide 2c",
Text: "Your AI recommended X. Can you reconstruct why it made that recommendation? For an auditor who's asking tough questions?",
},
{
Number: 6,
Slug: "slide2-key",
Step: "Slide 2 Key",
Text: "Black box is a documented rejection reason at the FDA. And traditional databases? They overwrite history. Every update erases what came before.",
},
// SOLUTION - Introducing StemeDB
{
Number: 7,
Slug: "slide3",
Step: "Slide 3",
Text: "StemeDB is a knowledge graph that stores claims, not facts. It's append-only. It's auditable. And it's built specifically for regulated industries where the stakes are high.",
},
// APPROACH - How it works
{
Number: 8,
Slug: "slide4a",
Step: "Slide 4a",
Text: "When your sources disagree, you actually see that disagreement. It's not hidden. It's not swept under the rug. It's visible.",
},
{
Number: 9,
Slug: "slide4b",
Step: "Slide 4b",
Text: "When a source gets retracted, you know exactly what's affected. In seconds. Not days of manual investigation.",
},
{
Number: 10,
Slug: "slide4c",
Step: "Slide 4c",
Text: "And history is preserved. Nothing gets silently overwritten. Ever.",
},
// CAPABILITIES
{
Number: 11,
Slug: "slide5a",
Step: "Slide 5a",
Text: "Let me walk you through what this enables. First, conflict visibility. You see when sources disagree, with confidence scores that tell you how much.",
},
{
Number: 12,
Slug: "slide5b",
Step: "Slide 5b",
Text: "Second, cascade invalidation. You retract a source, and instantly see every downstream decision it affected.",
},
{
Number: 13,
Slug: "slide5c",
Step: "Slide 5c",
Text: "Third, a complete audit trail. Every query is logged with full provenance. Ready to export for regulators.",
},
{
Number: 14,
Slug: "slide5-reveal",
Step: "Slide 5 Reveal",
Text: "And time-travel queries. You can ask: what did we believe on January first? And get the exact answer from that point in time.",
},
// SOCIAL PROOF
{
Number: 15,
Slug: "slide6",
Step: "Slide 6",
Text: "The FDA has now authorized over twelve hundred AI-enabled devices. Every single one of them requires an audit trail. Let me show you what compliance actually looks like.",
},
// DEMO - Conflict Visibility
{
Number: 16,
Slug: "demo1a",
Step: "Demo 1a",
Text: "I'm querying semaglutide gastroparesis risk. Notice the status says Contested. This immediately tells your analyst there's no clean answer here. Different sources are saying different things.",
},
{
Number: 17,
Slug: "demo1b",
Step: "Demo 1b",
Text: "Look at the weight distribution. FDA clinical trial data says 0.2 percent incidence. But patient reports are saying something different. Both are visible. Both have sources you can trace.",
},
{
Number: 18,
Slug: "demo1c",
Step: "Demo 1c",
Text: "Most databases would give you the FDA number and call it done. We show you the disagreement. Your medical affairs team can investigate before it becomes a problem. Nobody gets blindsided.",
},
{
Number: 19,
Slug: "demo1-amaze",
Step: "Demo 1 Amaze",
Text: "This isn't a recommendation from a black box. This is a recommendation with a complete evidence chain that you can trace back to every source.",
},
// DEMO - Audit Trail
{
Number: 20,
Slug: "demo2a",
Step: "Demo 2a",
Text: "Every query. Every agent. Every decision. It's all logged. Click any entry and you see exactly which assertions contributed to that decision.",
},
{
Number: 21,
Slug: "demo2-amaze",
Step: "Demo 2 Amaze",
Text: "Audit response time drops dramatically. What used to require manual log archaeology is now a single click.",
},
// DEMO - Cascade Invalidation
{
Number: 22,
Slug: "demo3a",
Step: "Demo 3a",
Text: "Here's a real FDA label. Over a hundred assertions in the system cite it as a source. Now imagine: the agency updates this label tomorrow morning with new safety data. What do you do?",
},
{
Number: 23,
Slug: "demo3b",
Step: "Demo 3b",
Text: "A JAMA study found that devices cleared using predicates with recall history had six point four times higher risk of future Class I recalls. When you can't trace which sources supported which decisions, you inherit that risk silently.",
},
{
Number: 24,
Slug: "demo3c",
Step: "Demo 3c",
Text: "One click. Preview Impact. Here's every decision that relied on this source. Your team can review them in priority order before anything goes wrong.",
},
{
Number: 25,
Slug: "demo3-amaze",
Step: "Demo 3 Amaze",
Text: "Time to identify impact goes from days to seconds.",
},
// DEMO - Time Travel
{
Number: 26,
Slug: "demo4",
Step: "Demo 4",
Text: "A patient had an adverse event eight months ago. Their attorney asks: what information was available to your system at that time? Can you reconstruct that state? We can. This is the exact state of the knowledge graph on that specific date.",
},
{
Number: 27,
Slug: "demo4-amaze",
Step: "Demo 4 Amaze",
Text: "Point-in-time reconstruction is native. It's not a manual archaeology project. It's a query parameter.",
},
// DEMO - Trust & Safety
{
Number: 28,
Slug: "demo5a",
Step: "Demo 5a",
Text: "What happens when things go wrong? Let's say someone tries to inject high-confidence assertions without proper credentials. A new agent claiming ninety-five percent confidence on a safety claim? That's suspicious. It goes to the review queue, not production. Humans decide.",
},
{
Number: 29,
Slug: "demo5-amaze",
Step: "Demo 5 Amaze",
Text: "Your knowledge base cannot be poisoned. And when something gets blocked, you know about it.",
},
// CLOSE
{
Number: 30,
Slug: "return-bridge",
Step: "Return Bridge",
Text: "That's the core of what StemeDB does. Conflict visibility. Cascade invalidation. Complete audit trails. Time travel. And trust controls. Questions?",
},
}