Coordination Audit — Every Inc (Every Media, Inc.)

Date: 2026-04-01

Organization Profile

Every is a ~20-person hybrid media-software-consulting company founded in 2020 by Dan Shipper. We publish a daily newsletter to 100K+ subscribers, run 4 AI software products (Spiral, Cora, Monologue, Sparkle), operate a seven-figure consulting practice for finance and tech firms, host the AI & I podcast, and open-source our compound engineering methodology (12K+ GitHub stars). Revenue: ~$1.2M ARR subscriptions (15% MoM growth) + $1-2M consulting, on <$2M raised.

Key structural feature: “Two-Slice Team” model — each product is run by a single GM using compound engineering (99% AI-written code). 3-person design team rotates across all products as an internal agency. Consulting led by Natalia Quintero. Editorial led by Kate Lee (EIC) and Eleanor Warnock (Managing Editor).

Time Allocation

Specification: 40% | Coordination: 27% | Execution: 33%

Assessment: Already remarkably close to AI-first target allocation (40-50% specification). Coordination at 27% is well below traditional orgs (~55%) but has 10-12 percentage points encodable. Most “execution” is human oversight of AI output — genuine manual production is minimal.

Workflow Analysis

Workflow 1: Article Production Pipeline

272 articles/year, 40 writers, 2-5 day turnaround from draft to publication.

Step Specification Coordination Execution
Pitch/ideation 70% 20% 10%
Writing (AI-assisted) 30% 10% 60%
AI tells detection 80% 10% 10%
Editorial review (3 rigor tests) 60% 30% 10%
Social distribution 20% 20% 60%
Workflow Total 45% 20% 35%

Key bottleneck: Kate Lee and Eleanor Warnock are throughput constraints for editorial review. AI tells detection (Katie Parrott) already partially encoded.

Workflow 2: Compound Engineering Feature Cycle

4 products, each run by 1 GM, Plan (40%) → Work (10%) → Review (40%) → Compound (10%).

Step Specification Coordination Execution
PRD/plan creation 75% 15% 10%
Agent execution 5% 5% 90%
14-agent code review 50% 20% 30%
Compounding/documentation 60% 20% 20%
Workflow Total 50% 15% 35%

Most AI-optimized workflow. 15% coordination is nearly optimal. Individual GM workflows differ significantly (Yash: parallel Claude+Codex, Danny: Droid CLI, Naveen: Linear-centric, Kieran: plan-first).

Workflow 3: Consulting Engagement Delivery

100+ companies engaged, ~two dozen deep. Finance and tech verticals. NPS >70.

Step Specification Coordination Execution
Client onboarding 30% 50% 20%
Workflow building 60% 20% 20%
Training sessions 40% 30% 30%
Ongoing “Chief AI Officer” support 30% 50% 20%
Workflow Total 40% 35% 25%

Highest coordination % of any workflow. Client-facing work inherently requires more coordination (calls, alignment, status updates). Natalia’s “Claudie” agent already saves 14 hrs/week on onboarding and weekly updates.

Workflow 4: Podcast Production (AI & I)

41 episodes/year, 83,957 listening hours. Rachel Braun leads.

Step Specification Coordination Execution
Topic/guest selection 60% 30% 10%
Recording 20% 30% 50%
Post-production (Descript) 15% 15% 70%
Distribution 20% 30% 50%
Workflow Total 30% 25% 45%

Most execution-heavy workflow. Post-production is heavily AI-assisted (Descript, StreamYard). Guest scheduling and distribution logistics are the main coordination costs.

Workflow 5: Product Design Rotation

3-person team (Lucas Crespo, Daniel Rodrigues, +1) serving 4 products + consulting + marketing.

Step Specification Coordination Execution
Request intake 30% 50% 20%
Design work (Figma) 50% 10% 40%
Handoff to GM (Figma MCP) 10% 60% 30%
Workflow Total 30% 40% 30%

Highest coordination overhead after consulting. The 3-person team context-switches across 4 products constantly. Request intake and handoff are coordination-heavy. Figma MCP integration already reduces some handoff friction.

Dual-System Classification

Structure Coordination Function Cultural Function Encoding Risk
Weekly all-hands/demo day Status alignment Core culture ritual — shared identity, “be sincere not serious” ethos, celebrating shipped work HIGH — never encode
Editorial standup Pipeline management, scheduling Writer community, editorial craft identity, taste culture MEDIUM — pipeline mgmt encodable; taste discussion is cultural
Product-design sync Resource allocation, priority Cross-team relationships, design team identity MEDIUM — scheduling encodable; design critique is cultural
Consulting pipeline review Client tracking, capacity planning Practitioner identity, consulting team culture MEDIUM — status tracking encodable; strategy is cultural
GM autonomy model Reduces coordination overhead Core identity — entrepreneurial ownership, two-slice team philosophy HIGH — protect fiercely
Slack async communication Status updates, quick questions Informal culture, community, belonging LOW — already async, can be structured further
14-agent code review Quality assurance, consistency Engineering craft, learning culture LOW — already gold-standard encoded
Editorial review chain Quality control, brand protection Editor authority, writer-editor relationship, taste identity MEDIUM — AI tells already encoded; final judgment is cultural
Linear for PM Task tracking, prioritization Individual GM ownership identity LOW — pure coordination tool
Compound engineering plugin Knowledge transfer, standards Open-source community, builder credibility LOW — already encoded and open-sourced

Encoding Candidates (Ranked by ROI)

1. Design Request Intake & Scheduling

  • Hours/month: ~40 (across all GMs + design team)
  • Coordination/cultural split: 80% / 20%
  • Encoding approach: Structured intake form in Linear with priority scoring + auto-scheduling based on design team capacity and product sprint cycles
  • Cultural risk: Loses informal GM-designer rapport. Mitigate: keep design critique meetings, encode only logistics.
  • ROI: HIGH

2. Consulting Client Status Reporting

  • Hours/month: ~30
  • Coordination/cultural split: 75% / 25%
  • Encoding approach: Extend Claudie (Natalia’s AI PM agent) to cover all routine status reporting; human touch reserved for strategy and relationship calls
  • Cultural risk: Client relationships feel less personal. Mitigate: human on all strategy calls, AI on logistics only.
  • ROI: HIGH

3. Editorial Pipeline Management

  • Hours/month: ~25
  • Coordination/cultural split: 70% / 30%
  • Encoding approach: Automated Kanban with AI-triggered nudges when articles stall, bottleneck data surfaced automatically, deadline tracking
  • Cultural risk: Loses organic editorial taste discussions during pipeline reviews. Mitigate: separate taste conversations from logistics conversations.
  • ROI: MEDIUM-HIGH

4. Cross-GM Knowledge Transfer

  • Hours/month: ~20 (measured as lost efficiency — each GM reinventing solutions)
  • Coordination/cultural split: 60% / 40%
  • Encoding approach: Extend compound engineering’s “compounding” step to auto-publish new docs/solutions/ entries to a shared knowledge feed. All GMs subscribe.
  • Cultural risk: Loses serendipitous peer learning moments. Mitigate: monthly “engineering show-and-tell” where GMs demo their best workflows.
  • ROI: MEDIUM

5. Podcast Production Logistics

  • Hours/month: ~15
  • Coordination/cultural split: 80% / 20%
  • Encoding approach: AI agent handles scheduling, generates show notes, auto-posts to distribution channels
  • Cultural risk: Minimal — logistics don’t carry cultural weight.
  • ROI: MEDIUM

Quick Wins

  1. Formalize design request intake in Linear — simple ticket template with priority, deadline, product context. Eliminates ad-hoc Slack requests. One day to implement.
  2. Extend Claudie to all consulting client status reporting — proven by Natalia. Expand scope to all active engagements.
  3. Create shared “learnings feed” — when any GM’s compound engineering loop produces new docs/solutions/ entries, auto-post to #engineering-learnings Slack channel. Zero new meetings, maximum knowledge transfer.

Cultural Red Flags

  • Weekly all-hands/demo day — The beating heart of Every’s culture. “Be sincere, not serious.” Never encode, never skip, never make it optional.
  • Editorial taste discussions — Conversations where Kate and Eleanor debate whether a piece “sounds authentically like the writer” ARE the editorial culture. Encode mechanics (AI tells detection), protect judgment (human review).
  • GM autonomy model — Any encoding that constrains GM autonomy threatens Every’s core identity. Encode coordination FOR GMs (shared tools, templates, knowledge feeds) not coordination OF GMs (mandatory processes, standardized workflows).
  • Builder credibility culture — Every’s edge is “practitioners who build with AI daily, not management consultants.” Any formalization that creates bureaucratic overhead contradicts this identity.

org-genome-builder — Every’s identity is strong but implicit. It lives in Dan’s articles, CLAUDE.md files, individual team habits, and shared intuition. Encoding it formally into a structured genome will:

  1. Enable consistent governance as agent usage scales (Plus One, compound engineering, editorial AI)
  2. Provide quality standards that new hires and consulting clients can reference
  3. Unlock downstream skills (quality gates, specifications, agent configs) that are consistent with who Every actually is
  4. Preserve what makes Every “Every” as the team grows beyond 20 people