A Weekly Brief of Only What's Actually New.
How we built a five-task AI pipeline that monitors YouTube creator content, scores it against what Automaton already knows, and delivers a weekly Slack brief of genuinely novel approaches — at roughly $15–30/month.
The Problem
Staying current on AI and automation creator content meant manually scrolling feeds, watching videos, and reading transcripts — most of which covered approaches the team already uses. The signal-to-noise ratio was terrible, and the time cost was high.
The problem wasn't access to information — it was filtering. Without a structured way to score new content against existing capabilities, everything looked potentially relevant and nothing was clearly worth acting on.
What We Built
We designed a five-task pipeline: discovery (surface new videos from tracked creators), transcript capture (extract and clean full transcripts), Claude analysis (score each video for novelty and relevance against the knowledge base), funnel capture (tag and store high-scoring content), and weekly brief (synthesize the week's genuinely new approaches into a Slack post).
The key was the knowledge-base table — seeded with Automaton's existing capabilities — that gives Claude a reference point for scoring novelty. A video explaining RAG is only novel if Automaton doesn't already use RAG. The scoring layer makes that distinction automatically.
“Staying current on AI creator content meant scrolling endless feeds and re-reading ideas we already operate on. The pipeline surfaces only what's genuinely new — filtered through what we actually do.”
The System Architecture
Five-task pipeline: YouTube Data API for discovery, transcript capture and cleaning layer, Claude analysis with knowledge-base scoring, Supabase for storage and knowledge base, Slack webhook for weekly brief delivery. Knowledge base seeded with existing Automaton capabilities. Novelty and relevance scores stored per video for trend analysis.
The Results
A weekly Slack brief that surfaces only genuinely new approaches — filtered through what the business actually does — at roughly $15–30/month in API and infrastructure costs.
The pipeline also created a structured knowledge base of Automaton capabilities as a side effect — a useful artifact for onboarding, positioning, and capability gap analysis.