Working glossary
A shared vocabulary for the Cosmo redesign — products, roles, entities, statuses, metrics. This is a working draft. If you spot a term that's used differently in your part of the org, flag it so we can unify before it ships.
Open question — Idea vs Task vs Prompt. The product currently uses idea in the Cosmo UI ("Ideas pending", "Ideas ready"). In Polo (and in some Apollo conversations), the same artifact is called a task. Prompt is sometimes used too, especially when referring to what gets sent to Apollo.
We need to pick one and be consistent. Default proposal: keep idea as the conceptual seed (Atlas-side noun) and task as the unit of auditor work (Apollo / Polo noun). If you disagree, ping Ishai.
Products / squads
- Cosmo
- Pipeline orchestration tool. Single source of truth for the production pipeline. Ishai's redesign target.
- Atlas
- Ideation engine. Generates prompts for playables.
- Apollo
- AI copilot for playable production. Auditor + AI agent build the playable together.
- Hermes
- Operational squad (HTML5 devs, asset manager, auditors, ops manager). The production engine.
- Helix
- Zero-shot base playable creation. Research → productionizing in 2026.
- Astra
- Video ad production (declining, ~60% of revenue).
- Prometheus
- Customer-facing Slack bot. Sett's first client product.
- McSett
- Future client platform / portal. Cosmo nav has a placeholder.
- Polo
- The auditor's task-management surface (sister product to Cosmo). Where auditors actually work on tasks.
Roles
- TL
Team Leader - Auditor team lead. Owns weekly capacity allocation. E.g., Tatiana, Daria, Aleksa, Mykyta.
- Auditor
- The person doing the actual playable work — typically using Apollo (and now Polo). Reports into a TL.
- PG
Partnership Growth - Client-facing role. Creates prompts, sends them to clients, collects revisions. Sometimes "Partnership Growth Manager".
- TPM
Technical Project Manager - Ops side, e.g. Einat.
- PM
Product Manager - E.g., Sergey.
- Squad lead
- Leads one of Atlas / Apollo / Hermes / Cosmo.
Entities
- Playable
- An HTML5 ad unit. The fundamental work item the pipeline produces.
- Game
- A customer game (e.g. Solitaire Cash, Bubble Cash, Woodoku Blast). One game has many playables.
- Concept
- A creative thesis behind a playable. Lives in Atlas.
- New Concept (NC)
- A playable based on concepts borrowed from other games (with a concept tag).
- Version (V)
- A playable created from existing playable versions of the same game.
- Revision (R)
- A playable that came back from the client with feedback. The same playable on another pass.
- Round
- Each revision counts as a round (R2, R3, …). Repeat revisions are the killer quality signal.
- Build
- The compiled output of a playable. Can fail.
- Core
- The base/template a game's playables are built from. "Pre-core" games haven't completed their first core yet.
- Base ≈ Core
- Used interchangeably with Core in many conversations.
- Idea also: Task, Prompt
- The creative seed that gets attached to a playable slot. Each new V slot needs 3 ideas in the game's pool before an auditor can start. Terminology in flux — see the callout at the top.
- Prompt
- The instruction sent to Apollo (or generated by Atlas) for the auditor to work from. Often used interchangeably with "idea" when in the auditor-facing Apollo surface.
- Quota commitment
- A planned playable slot for a specific game on a specific week. Sits in the Quota Backlog until a TL drags it onto an auditor lane. Three states: Ideas ready, Ideas pending, Ideas pending — nudged.
- Slot ≈ Quota commitment
- Shorter form of "quota commitment" — the placeholder card you drag from the backlog to a lane.
Statuses
Canonical 13 (playable lifecycle, Cosmo board)
Assigned
In Progress
Pending TL
Pending PG
Waiting for fix
Approved
Stuck
Build Queued
Building
Build Complete
Build Failed
Delivered
Completed
Slot readiness (pre-work, before status applies)
Ideas ready
Ideas pending
Waiting on PG
Game-level health (Status / Scorecard pages)
On Track
At Risk
Behind
Not Started
Metrics
- Quota
- Monthly playable target per game.
- Pace
- Daily delivery rate vs. required rate to hit monthly quota.
- Revision Rate
- % of versions that became revisions. Q1 target: 15%. Currently ~21%.
- Repeat Revision Rate
- % of revisions that needed multiple rounds. The killer quality signal.
- Avg Resolution
- Days from revision creation to resolution.
- SOV
Share of Voice - Share of voice in an ad channel. Winner = >20% SOV in top channel for 7 days.
- NRR
Net Revenue Retention - Target: 175% per logo / 6 months.
- ARR
Annual Recurring Revenue - 2025: $9M. 2026 target: $40M.
- Last mile
- Time the auditor spends finalising a playable.
Playable complexity tiers
- Level 1
- Simplest. Current Apollo focus.
- Level 2
- Slightly more complex. Current Apollo focus.
- Level 2.5
- Intermediate 3D. Helix Q1 target.
- Level 3
- Klondike-like complexity. 2026 expansion bet.
Workflow concepts
- Demand chain
- Weekly → Ideation → Atlas. The chain that produces the ideas that fill quota slots.
- Winner Search Loop
- Ideation → Production → Deployment → Data → repeat.
- Lightspeed
- Project Lightspeed (Jan 2026). Apollo velocity push.
- Popcorn Review
- Atlas async batch generation overnight; PG reviews like a queue.
- Project Lightspeed V1
- Q1 2026 Apollo initiative.
- Pulse
- The TL observability strip at the top of the Weekly screen — risk / status / trend modes.
Cultural
- Velocity Over Everything
- Sett's cultural mandate.
- Vibe-prototype
- Build a "rough-but-real" version that demonstrates core value. Velocity-way alternative to PRDs.
- No-Blocker Playbook
- The cultural ritual: don't wait for process, act directly.