Skip to content

How It Works

This repo is a multi-phase pipeline that turns a date window into a newsletter draft, with a paper trail in workspace/ that shows what was collected and curated along the way.

If you want to run it, start with: - Start here (Feb 2026)

Pipeline Phases

  1. Phase 1A: URL manifest generation
  2. Phase 1B: Source content retrieval
  3. Phase 1C: Discovery consolidation
  4. Phase 1.5: Curator notes (optional, if notes exist)
  5. Phase 2: Events extraction
  6. Phase 3: Content curation
  7. Phase 4: Newsletter assembly
  8. Phase 4.5: Polishing + validation

Phase Gates

Each phase writes explicit artifacts to disk and must pass existence/quality gates before the next phase.

Examples: - workspace/newsletter_phase1a_url_manifest_*.md - workspace/newsletter_phase1b_interim_*_*.md - workspace/newsletter_phase1a_discoveries_*.md - workspace/newsletter_phase2_events_*.md - workspace/newsletter_phase3_curated_sections_*.md - output/YYYY-MM_month_newsletter.md

Self-Learning Loop

Corrections are encoded into skills and references, then validated by scoring and rule checks.

Loop:

Finding -> Root cause -> Skill/rule update -> Regenerate -> Validate -> Record learning

Trust Disk, Not Self-Reports

The system relies on file-based proof and deterministic checks rather than agent claims.