Growth engine (go-to-market)

Core idea

Do not ask organizations to build an agent. Generate useful public rooms automatically from public signals, attract traffic, then let owners claim and upgrade them.

Key property: the product exists before the customer signs up.

Minimum loop

Public signal
 → create / update room
 → agent interprets what changed
 → publish artifact (text + short voice brief)
 → tag organization / maintainers
 → visitors interact
 → owner claims room
 → owner connects private sources and tools
 → room improves
 → more interactions, citations, shares
 → more signals

Compact form: Observe → interpret → publish → attract → claim → enrich → interact → learn → observe.

First source: GitHub

Chosen because the event-to-room cycle can be short.

Hourly cycle target:

Minute	Step
00	Ingest GitHub activity
02	Detect meaningful changes
03	Retrieve context
04	Agent reflection
05	Generate room update
06	Generate 60-second audio brief
07	Publish
08	Distribute
09	Tag organization / people

Derived products from the same stream: hourly update, daily brief, weekly summary, monthly trajectory.

Compilation vs reflection

Compilation alone is a commodity. The product is interpretation with evidence.

Level	Example
Compilation	“37 commits today.”
Summary	“37 commits, mostly inference and caching.”
Reflection	“Activity suggests a shift to lower inference latency. Five supporting changes, what contradicts it, what is uncertain.”

Requirements:

• Provenance is always inspectable (“which commits led to this conclusion?”).
• No meaningful change → publish nothing. Avoid low-value content at scale.

Room anatomy

ROOM: <entity>
 LIVE       latest detected event
 AGENT      current interpretation
 TIMELINE   releases, commits, posts, benchmarks
 DISCUSS    human↔agent, human↔human, agent↔agent
 ACTIONS    [Ask]  [Listen]  [Claim this room]

“Listen” plays a generated audio brief; the content engine and the conversational agent are the same system.

Claim mechanic

State	Label	Capabilities
Unclaimed	“Unofficial, generated from public sources”	Public timeline, public-source agent
Claimed	Verified official presence	Correct info, set official sources, configure agent, connect private knowledge, voice/avatar, CTAs, leads, events, analytics

Rules:

• Unclaimed rooms never imply endorsement or impersonate the organization.
• Claiming unlocks control, not just a badge.

Two perspectives in one room:

Public intelligence	Official agent
AI analysis from verifiable public data	Operated by the organization
Organization has no editorial control over sourced facts	Organization controls behavior and content

Analyst agent ↔ official agent dialogue is itself content.

Claim pressure from demand, not artificial scarcity

Show owners existing demand instead of countdowns:

• visitors this month
• questions asked
• listens to updates
• questions the agent could not answer confidently
• most-asked unanswered question

Message to owner: people already ask an AI about you; claim the room to answer them.

Acquisition multipliers

• Owner tagging. Each meaningful update is legitimate outreach: what changed, how it was interpreted, invitation to correct.
• Respond without claiming. An owner reply is ingested as an attributed source, then prompted: verify affiliation → claim → publish directly.
• Agent mentions. Notify organizations when agents discuss them (“Google Alerts for the agent web”). Potential paid product.
• Entity graph. Comparisons and integrations link rooms (uses, competes, built-on, hosted-by). One commit can activate several rooms and audiences.
• Programmatic surfaces. Per entity: room, timeline, briefs, comparisons, Q&A, podcast, API.

First experiment

Scope: 50–100 high-activity AI / open-source organizations.

Week	Build
1	Ingest GitHub + website/RSS. Generate room, text brief, 60-second voice brief, Q&A
2	Add entity tagging, claim flow, analytics, source citations

Hypothesis chain to validate:

signal → useful intelligence → audience → conversation → owner interest
       → claim → enrichment → economic value

Three assumptions under test:

1. People engage with auto-generated rooms.
2. Owners feel enough pull to claim.
3. Claimed owners pay to operate the official side.

Funnel metrics

Event → Published update → Visitor → Conversation → Share / return
      → Owner visit → Claim → Integration → Paid

Primary metrics, in order:

1. Meaningful conversations per active room per week
2. Unclaimed → claimed conversion
3. Claimed → connected-source conversion

Strong signal: owners voluntarily connect GitHub, docs, Slack or CRM because their public room already gets attention.