Xandrov StudioAtlas of shipped systems
← The territory/ II. Applied & Vertical AI/ № 08 — Signal IQ
Dossier № 08 / 19 — Applied & Vertical AI — Applied AI · Intelligence

Signal IQ

Competitive intel as a pipeline, not a panicked deck.

File
№ 08 / 19
Territory
Applied & Vertical AI
Classification
Applied AI · Intelligence
Status
Working system · Packaged offer
§ 01 Context

Competitive intelligence in most organizations means a single spreadsheet — created in haste, obsolete by board meeting day. The work is episodic and reactive: a scramble before investor updates, a panic before critical decisions. Signal IQ reframes the problem as a pipeline instead. Continuous ingestion of signals. Synthesis without urgency. Intelligence that compounds over time. The system tracks competitors and market moves as an ongoing workstream, feeding a durable intelligence brief rather than a one-off report. The value lives in accumulation, not in any single signal.

§ 02 The problem

The hard part isn't capturing any single piece of data — it's accumulation and synthesis over time. A scraped news item alone has limited value. A quarterly trend, observed continuously, means something. This demands infrastructure: scheduled background workers that run on a reliable cadence, a verifiable queue (because a silently dead worker is worse than one that fails loudly), and an LLM layer that can summarize signal streams without flattening their nuance. The question isn't 'can we scrape competitors?' It's 'can we trust the pipeline, and can we meaningfully compress what we see into an ongoing brief?'

§ 03 The approach

The system is built on Node and Express, organized around workstreams — the competitive domain (a company, a product category, a market move) is the first-class unit, not a generic feed. Clerk handles authentication. Postgres powers production; SQLite offers a local development path. Scheduled background workers ingest signals on a reliable cadence, depositing work into a persistent queue. A queue-verification script runs independently, auditing the queue state and surfacing stuck jobs before they go silent — loudness is a feature. The Gemini synthesis layer reads accumulated signals from each workstream and maintains the living brief, condensing trend and detail into actionable context. CLI tools (new, list, merge) manage the workstream lifecycle and queue inspection. The architecture treats background work as a first-class concern, not an afterthought.

The value isn't any single scrape — it's accumulation and synthesis over time.

§ 04 Decisions that mattered
D1
Workstreams as the unit
Intelligence has a subject, not just a feed. Each workstream tracks one competitive domain, accumulating signals and maintaining synthesis over time.
D2
Queue verification before silence
A background worker that fails silently is worse than one that fails loudly. The verification script audits the queue and surfaces stuck jobs.
D3
Synthesis layer with Gemini
The LLM condenses signal streams into an ongoing brief without flattening context. Nuance in accumulated signals translates to actionable intelligence.
D4
CLI for workstream lifecycle
Commands (new, list, merge) manage the workstream state. Queue inspection is a first-class CLI concern, not buried in logs.
§ 05 Where it stands

The system is in production and working. Competitor data that was scraped as part of the build is removed from anything published, as is any client-specific signal data. The studio offers Signal IQ as a packaged system and implementation — the architecture and operations model are stable. It is positioned as the ongoing alternative to spreadsheet-and-scramble, for teams that need to act on market moves before they calcify into industry consensus.

Node + ExpressClerkPostgresWorker + queueGemini
§ 06 — Sanitization record

Scraped competitor signals and client-specific data have been removed from the published system. We've withheld prospect lists, account identifiers, and any proprietary scraping logic. This is deliberate: the signal-pipeline pattern is the portable insight. Actual competitors, actual client contexts, and actual signal data are sensitive goods and stay behind the studio's lines.

Have a system like this worth building?

Fixed scope, quoted after one working session, first output inside two weeks. Read by a person within one working day.