SecurityV0 — How We Work

Operating plan for a 3-person AI-native security startup. v3 — 2026-03-21 — Revised after Owen review + Ivan decisions

Team

Role	Person	Focus
CEO	Sergey	Vision, GTM, partnerships, fundraising
CTO	Ivan	Architecture, implementation, connectors, infrastructure
Developer	(Dev)	Feature work, bug fixes, testing

1. The Idea Funnel

Every feature, research topic, integration idea, or customer request follows one path:

💡 CAPTURE → 🔬 RESEARCH → ✅ VALIDATE → 🏗️ BUILD → 🚀 SHIP

1.1 Capture (💡)

Goal: Nothing gets lost. Every idea gets written down within minutes.

Sources:

• Phone calls / meetings → Granola AI auto-captures → MCP extracts action items & ideas → Notion

• Slack conversations → manual or bot-assisted capture

• Random thoughts → Telegram to Lobster ("track idea: X") or direct Notion quick-add

• Customer feedback → tagged in Notion

Where it lands: Notion database SV0 Ideas

Notion DB schema:

Property	Type	Values
Title	title	Idea name
Status	select	💡 Idea / 🔬 Research / ✅ Approved / 🏗️ Building / 🚀 Shipped / ❌ Killed
Owner	person	CEO / Ivan / Dev
Source	select	Brainstorm / Meeting / Customer / UX-Review / Research / Slack
Category	multi-select	Platform / Connectors / UI / Infra / GTM
Notes	text	Decision rationale, links, research doc links, context
Meeting Link	url	Granola meeting link (if from a call)

7 properties. 6 statuses. This single database covers ideas, feature requests, AND feedback (feedback = idea with Source: UX-Review). No separate Feedback Tracker needed.

1.2 Research (🔬)

Goal: Validate feasibility and value before committing to build.

Who researches:

• Technical feasibility → Ivan or AI agent (Owen/Lobster)

• Market/competitive → CEO or AI agent

• Customer demand → CEO

Where research content lives:

• sv0-documentation repo: docs/architecture/research/ directory (existing, git-tracked)

• Existing lifecycle tags in frontmatter: idea → research-in-progress → research-complete → adopted/deferred/rejected

• Each research doc is linked from the Notion idea card (Notes field)

Status mapping — Notion ↔ sv0-documentation:

Notion Status	Repo Frontmatter	Notes
💡 Idea	idea	Just captured, no research yet
🔬 Research	research-in-progress	Someone is actively investigating
✅ Approved	research-complete → adopted	Decision made, moving to PRD/build
🏗️ Building	(no repo equivalent)	Tracked in GitHub Issues
🚀 Shipped	(no repo equivalent)	Done
❌ Killed	deferred or rejected	Won't do (now or ever)

Rule: Notion tracks status. sv0-documentation holds content. No duplication. Update BOTH when status changes (Notion card + research doc frontmatter).

1.3 Validate (✅)

Goal: CEO + CTO decide: build or kill.

Process:

1. Research is done → owner updates Notion card to ✅ Approved or ❌ Killed

2. Decision rationale goes in Notes field

3. Approved ideas get a PRD written as page content on the Notion card

Decision criteria:

• Customer impact (who asked, how many would benefit)

• Strategic alignment (does it advance the platform story)

• Effort vs. payoff

• Technical debt implications

1.4 PRD & Build (🏗️)

PRDs live in Notion. On the approved idea card itself — the card becomes the PRD.

PRD template (minimal, as page content):

## Problem
What's broken or missing?

## Solution
What are we building?

## Success Criteria
How do we know it's done?

## Scope
IN and OUT

## Technical Approach
Architecture notes, data model changes, API changes

## Tasks
High-level breakdown → becomes GitHub Issues

Implementation tracking: GitHub Issues on the relevant repo

• Each issue references the Notion card

• Labels: priority/P0-P3, type/feature, type/bug, type/infra, type/docs

• Milestones for larger features

1.5 Ship (🚀)

• PR merged → deploy to dev-sv0.fofanov.ai → verify → production

• Update Notion card → 🚀 Shipped

• Update docs in sv0-documentation if needed

2. Meeting Notes & Knowledge Capture

2.1 Granola AI

Status: Business subscription available via Lenny's Product Pass

Setup:

• Granola runs on Ivan's Mac during all SV0-related calls

• Auto-transcribes, generates structured notes with action items

• Official Granola MCP connects to AI agents (released Feb 2026)

Post-meeting flow:

Call ends
  → Granola generates summary + action items
  → Agent queries Granola MCP:
     - Extract ideas → create Notion cards in SV0 Ideas
     - Extract action items → add to existing Notion cards or create GitHub Issues
     - Extract decisions → add to Notes on relevant cards
  → Meeting summary saved as Notion page (linked from idea cards)

MCP options:

1. Official Granola MCP — hosted, connects to Claude/ChatGPT directly

2. Community granola-mcp (github.com/chrisguillory/granola-mcp) — local, reads cache, search + download

2.2 Ad-hoc Conversations

• Slack threads → flag important decisions for capture

• Telegram voice notes → agent transcribes, extracts action items

• Quick calls → enable Granola for all calls, not just scheduled ones

3. Weekly Rhythm

One ritual to start: Monday sync.

What	When	Who
Weekly sync	Monday	CEO + Ivan

Agenda:

1. Review Notion Ideas backlog — any new ideas to discuss?

2. Status check on 🏗️ Building items — blocked?

3. Assign owners for 🔬 Research items

4. Quick priority check — still working on the right things?

Add more rituals only after the Monday sync runs consistently for 4+ weeks.

Agent support: Lobster sends a morning briefing (cron, already configured) with task status, calendar, and email summary.

4. The Developer

The most actionable improvement in this plan is unblocking the developer:

1. Clear, well-scoped issues — no ambiguity about what "done" looks like

2. PRD context — every issue links to the Notion card with the why

3. Unblocked path — if blocked, it surfaces at Monday sync

4. Agent pairing — give them access to a coding agent for implementation help

5. Small wins — assign XS/S tasks first, build momentum

Onboarding the developer into this workflow is the highest-ROI action.

5. Documentation

sv0-documentation repo (existing, unchanged)

sv0-documentation/
├── docs/
│   └── architecture/
│       └── research/       ← Research docs (idea validation content)
│           └── YYYY-MM-DD-topic.md
├── connectors/             ← Connector specs
└── ...

Rule: Research content lives in git (sv0-documentation). Status tracking lives in Notion. Link between them via Notes field on the Notion card.

PRDs stay in Notion only. Research docs stay in git only. No sync between them — just cross-references (URL links in Notion Notes field, Notion URL in research doc frontmatter).

New directories (meetings/, decisions/) will be added only when there's actual content to put in them.

6. Tool Stack

Function	Tool	Notes
Idea tracking + PRDs	Notion (SV0 Ideas DB)	Single source of truth
Implementation tracking	GitHub Issues + Milestones	Per-repo, linked to Notion
Meeting capture	Granola AI	Business sub via Lenny's Product Pass
Research content	sv0-documentation (git)	Markdown, Obsidian-compatible
Communication	Slack + Telegram	Slack for team, Telegram for Ivan ↔ agents
Agent assistance	Lobster (OpenClaw) + Owen (ClaudeClaw)	Research, extraction, tracking
Secrets	1Password (sv0-bots vault)	Existing
Learning resources	Lenny's Product Pass	Product management frameworks & tools

7. Implementation Plan

Phase 1: Foundation (This Week) — Manual First

• Create Notion SV0 Ideas database with the schema above

• Backfill existing ideas from sv0-documentation research docs, Slack, memory

• Set up Granola on Ivan's Mac (confirm it works with SV0 calls)

• Run first Monday sync using the Notion board

• Onboard developer: walk through the workflow, assign first scoped issues

• Execute Notion workspace migration (see Migration Plan)

Phase 2: Integration (Weeks 2-3) — Validate Habits

• Is the Monday sync actually happening? Is Notion staying updated?

• Install Granola MCP → test querying meeting notes from agent

• Test: meeting → Granola → agent extracts ideas → Notion cards

• Build Notion integration: Lobster can create/update idea cards from Telegram

Phase 3: Automation (Week 4+) — Only Automate What Works

• Only automate processes that have been running manually for 30+ days

• Post-meeting idea extraction pipeline (if Granola MCP is reliable)

• Monday digest automation (if manual digest proved valuable)

• Research assignment workflow: "Research idea #X" → agent

8. What We're NOT Doing (Yet)

• Linear (not worth adding another tool for 3 people)

• Google Docs for PRDs (Notion only)

• ADR directory (add when we have decisions to record)

• Mid-week check-ins or Friday reviews (add only after Monday sync is habit)

• Full automation of the pipeline (manual first, prove the workflow)

---

v3: Fixed research path (docs/architecture/research/), added Notion↔repo status mapping, unified database (SV0 Ideas subsumes feedback tracking via Source field), clarified PRD/research separation (no sync, cross-references only). Reviewed by Owen, Claude Code agent, and Ivan.