LeadGrow
Operating view · verified 8 August 2026

Weekly competitor monitoring status

What we watch, what changes, what happens next.

A one-page operating view of the three competitors in LeadGrow's weekly watch: six public surfaces across Oxygen Agent, ColdIQ, and Deepline — the exact URLs, the real Monday schedule, the evidence chain, and the human response after a material change.[G]

3watched companies
6public pages checked
2active weekly jobs
Mon · 13:00UTC schedule
The automation stops at delivery. Owner, response SLA, ticketing, escalation, and strategic response are unverified or not automated today.[4]
Section 01

Who and what we watch

AI-native GTM workspace

Oxygen Agent

The watch covers Oxygen's homepage positioning, pricing structure, product claims, integrations, and agent compatibility across three public surfaces. The /product path returns a Clerk login wall — watched for structural changes to the gate itself.[G]

Unified GTM data API

ColdIQ

The watch covers ColdIQ's homepage positioning, provider count (40+), pricing tiers, agent platform support, and API/MCP claims.[G]

GTM SDK, CLI, and API

Deepline

The watch covers Deepline's homepage and quickstart: positioning, provider count (85+, 2,200+ tools), waterfall enrichment model, pricing, setup flow, and supported agents.[G]

Section 02

When it runs and where results go

There are two active jobs. Both use the cron 0 13 * * 1, so they run Mondays at 13:00 UTC; the delivery paths are different.[2]

CoverageJobPersonaScheduleDeliveryStatus
Oxygen Agent7edc0d99hephaestusMonday
13:00 UTC		Slack
Push delivery		Active[2]
ColdIQ + Deeplineb4ad9b2eirisMonday
13:00 UTC		Pantheon-only logging
No Slack push		Active[2]

Same cron, different visibility

Oxygen is pushed to Slack for immediate visibility. ColdIQ and Deepline are logged to Pantheon and require someone to look for the result.[2]

Observed firing order

The ColdIQ/Deepline job has appeared roughly six minutes after Oxygen despite the identical cron. This is recorded as scheduler queue ordering, not an intentional delay.[N]

Section 03

How changes become evidence

Detection is a semantic comparison: an LLM agent compares the meaning of extracted page content against the prior dated Markdown baseline, rather than comparing visual pixels or raw bytes.[3]

  1. ScrapeFetch every watched URL and extract Markdown.
  2. SnapshotSave a dated record in the competitor watch directory.
  3. BaselineLoad the most recent prior dated snapshot.
  4. CompareClassify material, minor, and unchanged content.
  5. DeliverSend to Slack or log to Pantheon.
Not usedHash or checksum diff
Not usedScreenshot comparison
Not usedDOM-selector comparison
Not usedRaw-HTML diff
01 · transientRaw page fetch

Web extraction output in the agent cache.

02 · canonicalDated Markdown snapshot

URLs checked, prior baseline, change summary, before/after detail, unchanged items, and assessment.

03 · interfaceDelivery record

Slack message for Oxygen or Pantheon entry for ColdIQ and Deepline.

  • Summary, not transcription. The agent's description is not a word-for-word diff; humans should inspect the dated snapshots for ground truth.[3]
  • No visual evidence. Pixel-only design, CSS, layout, and animation changes are invisible to Markdown extraction.[3]
  • Weekly granularity. A change published and reverted between Monday runs can be missed.[3]
  • URL maintenance. Moved or removed URLs can break comparison until the watched list is updated.[3]
  • Workspace-bound evidence. Snapshots are not themselves a shared public evidence archive; delivery is the team interface.[3]
  • Coverage gap. No prior ColdIQ/Deepline snapshot files or run evidence were found during grounding, although the scheduled job is active.[3]
Section 04

What people do next

  1. Read the delivery. Open the Oxygen Slack message or find the ColdIQ/Deepline Pantheon entry.
  2. Confirm materiality. Treat the agent classification as a prompt for judgment, not the final decision.
  3. Compare dated snapshots. Inspect the current and prior Markdown records directly.
  4. Choose a response. Update positioning, brief sales, adjust messaging, open a tracked issue, or explicitly take no action.
  5. Record the decision. Leave a durable Pantheon note, Paperclip issue, or Slack-thread reply.
GapCurrent stateWhat would close it
OwnershipUnassignedName a weekly reviewer.
SLANone verifiedDefine review and decision windows.
TicketingManualCreate a tracked issue when a material change is confirmed.
ColdIQ/Deepline visibilityPantheon-onlyAdd push delivery or a monitored dashboard surface.
ColdIQ/Deepline executionUnverifiedConfirm the job produces valid snapshots and delivery records.

Traceability

Source register

  1. [G] Verified ground truthdocs/forge/789ba472/GROUND.md — live page fetches and baseline facts (2026-08-08)
  2. [P] Synthesis requirementsdocs/forge/789ba472/PLAN.md
  3. [N] Cross-step constraintsdocs/forge/789ba472/NOTES.md
  4. [1] Who and what we watchdocs/competitor-monitoring/sections/01-who-and-what-we-watch.md — sourced from GROUND.md
  5. [2] Schedule and deliverydocs/competitor-monitoring/sections/02-when-it-runs-and-where-results-go.md
  6. [3] Change evidence modeldocs/competitor-monitoring/sections/03-how-changes-become-evidence.md
  7. [4] Human responsedocs/competitor-monitoring/sections/04-what-people-do-next.md