CI-OS Product Doctrine — Arijit, verbatim intent (2026-07-08 morning)

Recorded on Arijit's explicit order: "record all this shit... this is the gold." This document outranks any implementation detail. Every brief, dashboard, and module gets measured against THIS, not against what V0 did or what any competitor tool does.

The motto

"What is my competitor doing, and based on that, what intelligent task and step and work should I do?"

It is called INTELLIGENCE, not data mining. Data is mined every single day to BUILD intelligence.

The persona and the moment (the whole product in one scene)

I am the CMO of a company. I come in, I log in, and my CI-OS shows me, in very clear visual terms:

• "This is what YOUR company is doing."
• "This is what THAT competitor did."
• "This is what is relevant to you FOR TODAY, for you to take action and involve this team and that team."

Where your competitor is. Where you are. Where you think they could go.

Hard product rules (Arijit's words, normalized)

1. NOT Algolia-specific. Algolia is just the experiment currently running. CI-OS is a software product: modular, sellable to any company. Nothing in the intelligence layer may hardcode Algolia's vertical, competitors, or vocabulary.
2. No feeder screen. No onboarding questionnaire. "That's the 1950s." When CI-OS is sold to a company (target price point: ~$10,000), the seller has already researched the buyer's business and pre-built the competitor set. The customer logs in and it already knows who they are.
3. Completely autonomous. Monitor all competitors, check what they're doing online, grab the signals, transform signals into something that triggers an action for marketing (content or otherwise). No babysitting.
4. Own scouting capability. Learn by scouting the web ourselves: features, functionality, posts, everything, on a scripted 24-hour cycle.
5. The cadence ladder — mine daily, boil upward: - Daily (24h): digest everything learned, deliver succinct, lucid intelligence to marketing leadership: "In the last 24 hours, THIS happened; THIS is where you should pay attention." - Weekly: summarize, identify PATTERNS, provide a summarized action plan. - Monthly: boil the weekly up another level.
6. Every signal lands as an action naming which team gets involved (marketing, content, product, exec...). Signal without an action trigger is noise.
7. Business stakes: "based on these capabilities, our business, our product, is going to fail or bomb or be successful." This doctrine is the success criterion.

What this changes in the build (translation to work items)

• The daily brief template is written to the CMO persona: your-company vs competitor framing, today's attention items, team-level action triggers. Tenant = "your company"; the brief speaks to the reader as the leader of that company, never as an Algolia analyst.
• Weekly synthesis = pattern detection over the daily ledger + action plan. Monthly = roll-up of weeklies. Both are first-class pipeline outputs, not afterthoughts.
• The dashboard's first screen answers: where competitor is / where you are / where they could go — the "three positions" view.
• Tenant onboarding is a seller-side pre-build (competitor set researched and seeded before the customer ever logs in), not a customer-side wizard.
• Modularity: intelligence layer must be domain-agnostic; Algolia's competitor set is seed data, never logic.

Addendum (same day, later): freshness USP + Hermes packaging intent

The USP is FRESHNESS. Intelligence must be actionable AT THAT INSTANT — "even 30 minutes after it may not be applicable... half a day after it will not be applicable." The system monitors every channel and source, online and offline, gets it inside, analyzes, and hands over critical actionable instructions while they are still actionable. Daily cadence is the floor, not the ceiling; the architecture must support near-real-time signal-to-action.

Packaging intent: CI-OS ships WITH Hermes. Arijit wants Hermes and the custom Claude-built code (the V2 cios codebase) to become ONE product — "one common code base and one common spine of execution" — with CI-OS as Hermes's first flagship use case. Open architecture question, explicitly assigned as a goal/loop to figure out: can Hermes be the center (gateway, channels, scheduling, agent runtime) with cios as the domain brain, or do they merge differently? Study assigned 2026-07-08.

Addendum 2 (same day): why Argus beats a Klue subscription — the dot-connector

Arijit, verbatim intent: "If I'm a user and I got a Klue subscription, I got all the professional subscriptions — why the hell should I even look at Argus?" The answer, and therefore the product bar:

1. Argus connects the dots. Not another feed. It does research over the last 10 / 30 / 60 days (multi-horizon lookbacks) across the INDUSTRY, not just the named competitors, and brings in what is interesting and helpful.
2. Argus researches the tenant's OWN brand too. (Algolia today — plug and play, just the example.) Own-brand perception/position research is a first-class module beside competitor intel.
3. Contextualize → apply → prescribe. Gather intel, contextualize it to MY business, and hand me VERY specific strategies — marketing strategies, ploys — grounded in what was gathered. Not observations. Prescriptions.
4. Offload collaterals from the intelligence. A module that generates marketing collateral from gathered intel: marketing dashboards and marketing landing pages are the first two named. Intelligence that ends in a produced asset, not a paragraph.
5. Living doc discipline: Arijit adds product truth "as I'm remembering" — capture every drop here, keep enhancing. This doc is append-only gold.

The dashboard experience bar (Addendum 2 continued)

Arijit has not yet seen a screen he considers Argus. The old CI screen is his only reference. Dashboard must earn the "extraordinarily special" verdict: the three-position view, multi-horizon industry research, own-brand read, prescriptions with team routing, and generated collaterals — visible, premium, interactive. Design phase = frontend-builder flow against THIS list.

Addendum 3 (same day): multi-user, multi-channel, content engine, company-swap

1. Multi-user, multi-channel. Multiple people get Argus. Channels: Telegram, email, Slack, iMessage push — and the dashboard ITSELF is a channel that auto-updates with the daily focus and the weekly synthesis. Per-user, per-channel subscription is a product primitive, not a config hack.
2. Content engine. Argus recommends content: based on competitor moves, on what content OTHERS are putting out that is moving the market or sentiment, and on what the tenant should publish in response. (Schema already reserves content_recommendations + weekly_content_plan.)
3. Company-swap modularity (the product test). Algolia is just the test bed. The mechanisms, infrastructure, decision logic, execution rigor, interactivity, frequency, methodology, structure ALL stay constant; only the primary company swaps — and the competitor list + source list update as a consequence. Specifically: the SOURCE GATHERING module must be a kick-offable pipeline that re-runs whenever the central company changes (company in → competitors researched → sources discovered/validated/seeded). This is the seller-side pre-build from rule 2, made executable.