Signal ScoutSystem Brief
Open source v1.0
Open source market intelligence

An open-source tool for structuring signals into content and GTM media production.

Signal Scout scans the market for emerging topics, scores them, and routes the strongest through review. Users can then structure approved signals into outputs for content and GTM media production.

40+
signal sources mapped
7 wired at V1
12
scoring layers
weighted · multiplied · overridden
14
narrative frameworks
SPARK is primary
5
production formats
video · podcast · social
What it is

A market intelligence tool with a human in the loop

Signal Scout is a market intelligence tool. It watches a configurable set of external sources, identifies which topics are worth acting on, and passes the strongest to a person before anything is produced. Once a signal is approved, users can structure it into the components for content and GTM media production: talk tracks, beat maps, shot lists, storyboards, hooks, and post copy.

It runs through two paths. An automated pipeline scans sources on a schedule and scores each item before it surfaces. A manual pipeline lets a user drop in a note, file, or link and route it straight to output, skipping the scoring queue because a person has already vouched for it. Both end at the same step: choosing how to frame the signal and what to produce.

01IngestRSS, API, and search pull items from mapped sources on a cron schedule
02ScoreA 12-layer engine ranks each item on emergence, relevance, authority, and timing
03ReviewScored signals land in a queue for a person to approve, defer, or discard
04StructureA framework and one or more output formats are selected for the approved signal
05ProduceStructured outputs are generated for each selected format
The system, broken down

Two halves: getting signal in, and turning it into output

The pipeline is easiest to read as two connected halves. The first decides what is worth paying attention to. The second decides how to say it and what to make. Each has its own breakdown below.

Half one

Signal Ingestion

Ingestion is the market-intelligence half. Its job is to notice the right topics early and pass only the strongest ones forward, so that human attention is spent on decisions rather than scanning.

Sources & trust

Sources are tiered by how much the system trusts them. Research and academic feeds start high, trade press sits in the middle, and community sources start lower. New sources enter at neutral trust and earn or lose standing over time based on whether their signals get approved and whether outputs built from them perform.

0.85–0.95Research / academic  MIT Tech Review, arXiv
0.75–0.85Trade press  Wired, TechCrunch, VentureBeat
0.60–0.70Community  Hacker News, Reddit
0.50New source  neutral entry, earns trust over time
Three ingestion methods

The automated pipeline reaches sources three ways, each on its own cadence. Whatever the method, every item lands in the same shape and passes through scoring before a person ever sees it.

RSSFeed polling  every 30–60 min · podcasts, press, arXiv, Product Hunt
APIDirect  every 15–30 min · Reddit, YouTube, Google Trends, HN
ScrapeSearch & crawl  every 2–4 hr · web search, GitHub Trending
The scoring engine

A 12-layer engine scores how much a topic matters right now. Five layers feed a weighted base score, others apply multipliers, and two act as overrides: named voices auto-elevate, off-topic keywords zero out. Weights are configurable per deployment.

base = emergence 0.30 + relevance 0.25 + authority 0.20 + question gap 0.15 + velocity 0.10 → multipliers → overrides

All 12 scoring layersexpand
01Emergence Detectionweight 0.30
Concepts in the 5% window: past too-early, before mainstream saturation.
02Thought Leader Watchlistoverride
Named Tier 0 and Tier 1 voices auto-elevate above 0.90 and bypass the formula.
03Question Gap Detectorweight 0.15
Repeated unanswered questions across comments and threads signal an opening.
04Practitioner vs. Analyst Divergencefeeds emergence
Gap between analyst coverage and practitioner sentiment marks a positioning angle.
05Competitive Gap Intelligencefeeds relevance
What the major firms all cover, none cover, or cover poorly.
06Temporal & Calendar Intelligencemultiplier
Seasonal weighting for budget season, pre-conference, and year-end windows.
07Source Trustweight 0.20
Per-source trust that decays on low signal and grows on validated ones.
08Engagement Velocityweight 0.10
Comment volume, shares, and reactions measured at ingest time.
09Cross-Platform Heatmultiplier
Same topic across two or more platforms within 48 hours.
10Relevance Keywordsweight 0.25
Weighted keyword matching across AI/ML, data, enterprise, governance, and open source.
11Noise Filterzero-out
Off-topic keywords zero the score, so the item never surfaces.
12Gartner Hype Cycle Positionmultiplier
Position in the hype cycle sets a 0.7x to 1.5x multiplier.
The review queue

Scored signals land in a queue where a person makes the call. A threshold slider controls what surfaces, each card links back to its original source, and every approve or purge feeds back into source trust over time. This is the human-in-the-loop checkpoint.

> 0.85Immediate  alert, talk track within 24 hours
0.70–0.85Queued  enters the review queue
0.50–0.70Digest  lower priority, weekly roundup
< 0.50Logged  recorded but not surfaced
Half two

Structured Outputs

Once a signal is approved, users choose how to frame it and what to produce. Signal Scout structures the approved signal into per-beat outputs for each selected format.

Framework selection · optional

A framework sets the narrative structure and voice. SPARK is the primary one, and thirteen others cover short-form hooks through long-form explainers. Style modes and a set of anti-patterns keep outputs consistent and away from hype.

SPARKSignal · Position · Argument · Reinforcement · Kicker
+13PAS, BAB, StoryBrand, AIDA, Data-Driven, Hot Take, Story Arc, and more
Thought LeaderEducational ProvocativeCorporateConversational
Format selection · required, multi-select

Users can select several formats for one signal, and each produces its own structured output. Select talking-head and sizzle, and both are generated.

VideoShort-form, sizzle, long-form  script, beat map, shot list, storyboard
PodcastEpisodes  segments, guest prep, show notes, talking points
SocialPosts  platform copy, headline and hook variants, visual direction
ReviewBeats  every format opens as beats; each beat shows only the lanes that apply

Reviewed as beats. The review console opens each output as a sequence of beats. Each beat shows only the production lanes that apply to it: on-screen text, visual, camera, audio, voiceover, talking points, guest questions, and mood. A beat or document toggle switches between the per-beat view and the full write-up.

Beat labels and timestamps sit on each beat, live web sources sit under every output, and any output the engine stopped early is flagged.

Open the review console
Open source

Open source, configured per deployment

Signal Scout ships as open source with an empty or example configuration. Nothing is hardwired to a single industry. The core handles ingesting, scoring, queuing, and structuring; each deployment supplies the parts that make it specific.

What ships

A pip-installable Python package with the scoring engine, the ingestion connectors, the framework and format libraries, and default settings. Users clone it, configure it, and it runs end to end.

What each deployment adds

Its own source feeds, its own search queries and tracked communities, and its own scoring keywords. The same core tuned to enterprise AI and data surfaces different signal than one tuned to, say, consumer health.

How it's used

Two jobs: intelligence, then structured outputs

Signal Scout does two jobs. As market intelligence, it surfaces which topics in a domain are emerging before they saturate, where practitioner and analyst views diverge, and what questions keep going unanswered. Users can then structure those findings into outputs for content and GTM media production, so work starts from a scored signal and a structured brief.

Scoring is transparent and the review queue keeps a person in the loop, so it stays a market intelligence tool, not an automation that publishes on its own.