# Product Thesis

## User problem

Capture is abundant; attention is scarce. The primary user may save ten YouTube videos, several X posts, articles, and repositories in one day. A raw archive preserves too much friction. A fully automatic concept graph creates too much synthetic structure.

## Wedge

Wiki Engine adds a first-class **Signal** layer between Source and Wiki.

A good Signal note answers:

1. What is the bottom line?
2. Which insights are actually novel or useful?
3. What evidence/timecode supports each one?
4. What should I do differently?
5. Which linked resources deserve separate inspection?
6. What remains uncertain or transcript-dependent?

The daily success metric is not pages generated. It is attention returned without provenance lost.

## Lifecycle

- **Intake** registers exact source bytes.
- **Signal** extracts one source into a concise reading note.
- **Video Signal** escalates from transcript to targeted/full visuals only when justified.
- **Compile** proposes cross-source Wiki pages.
- **Review** surfaces missing work, duplicates, contradictions, and organization debt.

## Deliberate exclusions

The alpha has no daemon, feed scheduler, universal scraper, embedding database, autonomous promotion, 0–100 importance score, social posting, or hosted account. It does not turn every named entity into a page.

## Product standard

Wiki Engine should become as trustworthy operationally as Session Save: deterministic identity, portable records, safe installation, explicit source attribution, bounded reads, recoverable writes, backward-compatible schemas, adversarial tests, and honest public claims. Alpha.1 establishes the original kernel and workflow; crash-journaled cross-surface installation and broader platform witnesses remain future gates.