Signal Intake
Aggregates multi-source news, enriches full text, deduplicates, clusters, and ranks signals into a calm morning brief.
MORROW
Morrow
AI-Native Morning Intelligence Ritual
A native macOS news product for high-information-density users. Morrow upgrades news consumption into judgment acquisition: multi-source ingestion, deduplication, clustering, and ranking reduce noise first; then AI generates summaries, Why It Matters, multilingual renders, and event narratives; finally menu bar, Home, and Detail views complete a fast-scan to deep-read ritual.
8 min
Morning Loop
5 Lang
Languages
4 Layers
Abstraction
PRODUCT LAYERS
A Four-Layer Judgment Loop
Morrow upgrades news consumption into judgment acquisition: Signal Intake, Article Intelligence, Story Graph, and Ritual Surface.
Judgment Layer
Beyond summaries: generates Why It Matters, key points, and structured insight to accelerate prioritization.
Story Graph
Lifts Article into Situation, CanonicalEvent, and Narrative so users see what is happening now and what to keep watching.
Morning Ritual
Menu bar enables a 30-second glance, while the main window supports deep reading and review loops.
SYSTEM ARCHITECTURE
System Architecture
Client, backend, AI models, and knowledge layers work together to turn raw news into consumable judgment.
Client
SwiftUI / macOS Menu Bar + Main Window
Backend
Python / FastAPI / SQLite
AI Models
GPT-5.4 primary + Gemini / MiniMax fallback
Knowledge
Article -> Situation -> Event -> Narrative
AI TECHNIQUES
AI Techniques
A four-layer AI pipeline: Signal Intake, Article Intelligence, Story Graph, and Ritual Surface.
Layer 1: Signal Intake
Multi-source ingest + dedup + ranking
Fetch -> Normalize -> Deduplicate -> Cluster -> Rank. Raw RSS noise is compressed into a calm morning brief before anything enters the AI chain.
Layer 2: Article Intelligence
Summary + Why It Matters + 5-language render
Each article produces more than a summary: Why It Matters, key points, and five-language reading versions. Failed locale QA falls into deterministic repair plus AI repair.
Layer 3: Story Graph
Situation -> CanonicalEvent -> Narrative
The system does not stop at single-article summaries. It elevates news into scenes, canonical events, and evolving narratives, so the home view reflects world state rather than feed noise.
Layer 4: Ritual Surface
Menu Bar + Home + Detail loop
Menu bar handles the 30-second morning glance, Home supports fast scanning, and Detail combines Why It Matters, Situation Lens, and Narrative Arc into a full read-review loop.
DEEP DIVE
Deep Dive: Four-Layer Judgment Loop
The complete four-layer pipeline from news to judgment.
Morrow Intelligence Stack
A Four-Layer Judgment Loop
Signal Intake -> Article AI -> Story Graph -> Ritual Surface
Layer 1
Signal Intake
Multi-source ingest, full-text enrichment, deduplication, clustering, and priority ranking.
RSS / RSSHubFetch & Normalize
Dedup + ClusterDaily Brief Builder
Priority EngineRank / Personalize
Layer 2
Article Intelligence
Summary, Why It Matters, key points, and five-language reading renders in one chain.
Insight ChainSummary / Why It Matters
5-Language Renderen / zh / ja / ko / es
QA + RepairDeterministic Guard
Layer 3
Story Graph
Article -> Situation -> CanonicalEvent -> Narrative, turning scattered stories into world-state understanding.
Situation LensWho did what to whom
Canonical EventWhat is happening now
Narrative ArcWhat to keep watching
Layer 4
Ritual Surface
Menu Bar handles the 30-second glance, Home drives fast scanning, and Detail closes the loop with Why It Matters and deep reading.
Menu Bar Ritual30-second Morning Signal
Article DetailWhy It Matters / Linked Reads
Home WindowHome / Situations / Settings
30s
Menu Ritual
5 Lang
Localization
4 Layers
Judgment Stack