Signal Factory

Turn content noise into structured intelligence, automatically.

Define Sources

15 YouTube channels configured

Curate which YouTube creators to monitor. Each channel is tracked by its unique ID. New channels start producing intelligence on the next daily run.

Configure: Channels page

Ingest New Videos

YouTube Data API v3 · runs daily at 06:00 CET

Every morning, the pipeline queries each active channel for videos published in the last 24 hours. Duplicates are filtered against the Notion database. Only genuinely new content proceeds.

Analyze Each Video

Gemini 2.5 Flash · native video understanding

Each video URL is sent directly to Gemini — no transcript needed, it watches the video. Returns structured JSON: utility classification (TECHNICAL / STRATEGIC / WEAK_SIGNAL / NOISE), content type (GENUINELY_NEW / REHASH / CLICKBAIT), watch recommendation, thesis, practical findings, and tools mentioned.

Configure: Daily Extract prompt

Apply Analysis Lenses

2 configurable perspectives per video

Each video's analysis is evaluated through two lenses. The AI Builder lens scores relevance to system architecture, agent frameworks, and LLM APIs (1–10). The GTM lens scores relevance to agentic commerce — AI-driven sales and marketing automation (1–10). Both produce specific takeaways and action items.

Configure: 11 prompts configured

Store in Notion

One page per video · structured properties

Full analysis written to the Notion Videos database: title, creator, URL, publish date, utility label, watch recommendation, thesis, AI Builder relevance, GTM relevance. The page body contains the complete analysis breakdown. This becomes the source of truth for all downstream synthesis.

Synthesize & Deliver

Daily → Discord · Weekly/Monthly → Notion + Discord

Daily digest at 07:30 — featured videos posted to Discord, sorted by watch recommendation. Weekly (Sunday) and monthly (1st) — the pipeline pulls all video analyses from Notion, runs Perplexity web search for broader AI news, then synthesizes everything through Claude Haiku on OpenRouter into top signals, trend analysis, and narrative angles.

Configure: 4 AI models configured

OpenClaw News

Parallel track · daily at 17:00 + weekly on Fridays

Independent intelligence track monitoring the OpenClaw open-source AI agent framework. Three parallel searches (GitHub activity, community discussions, security advisories) are synthesized into a daily briefing with deduplication against yesterday's report. Weekly summary rolls up the week's developments with ecosystem context.

Content Production

Coming next

The signal inventory feeds a content production pipeline: LinkedIn posts crafted in a configured brand voice, published via Typefully, with generated artwork using OpenClaw templates. Transform intelligence consumption into thought leadership production.

Roadmap