Cosmo IA Restructure · M2

Cosmo Information Architecture (V2)

What stays, what goes, what is new. Plan for the Sunday walkthrough with Sergey and Lee. Not a final design, a structured conversation starter.

StatusDraft AudienceSergey + Lee WalkthroughSun 2026-05-10 OwnerIshai Revach

Today's Cosmo, classified

#sitemap

Every page that exists today. Classification is a conversation-starter, not locked. Five columns: what the page does, which bucket it belongs to, where it goes in V2, which jobs it serves, who it's primarily for.

Classification
Destination
Who it is for
Showing all 21 rows.
Page Classification V2 destination Jobs to be done Who it's for
In-nav today (11 items)
Status
Monthly quota progress per game and team. Currently BROKEN (P1, P34).
Observability
Drop nav
Functionality folds into Scorecard (M3) + per-role observability surface.
  • Track monthly delivery vs quota
  • Spot games behind pace
  • Status meeting prep
TPM CTO
Dashboard
Weekly delivery + ready-to-send queue. Misnamed, near-empty body.
Observability
Drop nav
Salvageable fragments go to Weekly, Scorecard, and PG Operational aging banner.
  • Get a TL daily-overview at a glance (no clear primary today, dead audience)
unclear
Ideation
TL workbench for sending tasks to Apollo + PG idea-pipeline + import. Manual Send-to-Apollo, multi-tab.
Operational Management
Split + drop
Operational buttons (Request Ideas, Add Prompt) go to PG Operational header. Numbers-view goes to M3. Import tab and Revisions go to Settings and Weekly Quota Backlog.
  • TL: Send a batch of prompts to Apollo (P50/P51/P52)
  • PG: Trigger Request Ideas
  • PG: Add Prompt to game program
  • Cross: Filter ideas-flow by PG/TL/Auditor/Game
TL primary PG secondary
Auditors
Per-auditor stage breakdowns. METRICS BROKEN (1,447 failed vs 2 done, P14, P30).
Observability
Drop nav
Allocation goes to Settings:Allocations. Per-auditor Monday-morning view goes to Weekly + Auditor Detail drawer. Scoreboard goes to M3 observability (Pulse).
  • TL (Tatiana P63): Per-auditor Monday-morning open-tasks view
  • TL: Auditor 1-on-1 prep
  • Head-of-PG: Auditor scoreboard
unclear
Builds
Active + failed builds list. No failure notifications today (P6, P13, P26).
Observability Management
Drop nav
Activity log goes to Management and Settings. History trends go to M3. Per-card chip + Build Detail drawer go to Weekly + PG Operational. Ambient notifications go to Slack.
  • TL: Investigate a stuck build
  • TL: See per-playable build status
  • TL/Auditor: Know when build fails and why
TL primary
Delivery
Ready-to-Send / Revisions / Waiting / Completed queues. Functional but low PG adoption (P28, P31, P32).
Operational
Replace + rename
Becomes PG Operational. Absorbs Ideation PG fragments + Game Page PG cuts + customer interaction log.
  • PG: Send a delivery to client
  • PG: Swap playable in brief (P44)
  • PG: Track per-game delivered-vs-quota
  • PG: Triage Need-Action queue
  • PG: Watch aging items
PG primary
Feedback
3,988 entries, zero aggregation today.
Observability
Drop nav
Browse-my-past-feedback goes to Management and Settings (Settings:Feedback-archive). Analytics half goes to M3 observability.
  • Latent: PG/TL "find feedback I gave"
  • Latent: PG/TL "track feedback I received"
  • M3 analytics: feedback trend per game / per role
unclear
Weekly
Mon-Fri kanban grid, auditors as rows, task cards in cells. V2 MVP.
Operational
Stay, main nav
V2 build target, ships M1. Absorbs Revisions backlog, Auditors Monday view, Status/Dashboard quota progress, Builds chip, Carryover.
  • TL: Plan the week
  • TL: Triage daily
  • TL: Send prompts batch to Apollo
  • TL: See carryover from prior weeks
  • TL: Reassign work mid-week
  • TL: Investigate stuck build
  • TL: Manage capacity per auditor
  • Cross: Recognize a playable instantly
TL primary PG secondary read
Revisions
Quality investigation tool with trend chart + repeat-revision highlighting. Sergey-era spec.
Observability Operational
Split
Allocation half (revisions backlog) goes to Weekly. Analytics half (trend, repeat highlighting, per-game grouping) goes to M3 observability (Pulse or standalone).
  • TL: Investigate revision causes
  • TL/PG: Spot repeat-revisioned playables
  • Head-of-PG: Get revision rate per game
  • TPM: Reliable revision rate / time-to-resolution / success %
TL primary PG secondary
Scorecard
Pace-based alerting per game. Best-designed surface today.
Observability
Stay, main nav
Kept as-is, M3 design-pattern source for the rest.
  • TPM/CTO: See live pace per game
  • TPM: Get reliable monthly delivery vs goal
  • CTO: Spot risk indicators across portfolio
  • Head-of-PG: Per-game trend
TPM primary CTO primary
McSett
Internal window into shipping autonomous fix-bot. Backend mcsett.py, 1,333 LOC, 8 endpoint groups, full pipeline write authority.
Observability
TBD nav, M3
Stay as own surface, fold into per-role observability, or graduate out of Cosmo when McSett-product matures. Audience decision needed.
  • Head-of-PG / TPM (TBD): Watch the autonomous bot's fix queue
  • Spot escalation patterns
  • Audit per-fix activity (timeline)
TBD
Drill-downs and off-nav (accessible but not in main nav)
Game Page
Per-game knowledge, delivery, tier, quota, customer link. Drill-down.
Operational Management
Stay, drill-down
Reachable from PG Operational + Scorecard. Per-role content split needed (PG vs TL vs TPM). Resolves P22.
  • PG: Manage customer-facing game
  • PG: Review brief + delivery routing per game
  • TL: See pipeline + capacity per game
  • TPM: Pace + trend per game
  • Yuval R2: Client-level cross-game roll-up
PG primary TL TPM
Admin
Allocation + Games + Admin panel sub-views. Off-nav menu today.
Management
Fold into Settings
Drops entirely into Management and Settings (Bucket C).
  • Admin: Edit entities (TL/PG/Auditor/Game)
  • Admin: Reallocate entity to entity
  • Admin: Adjust system preferences
Admin
AllocationPage
Entity-entity matrix (auditor-game today). Off-nav.
Management
Fold into Settings
Folds into Settings:Allocations.
  • Head-of-PG / Admin: Reassign auditor-game
  • View Game-TL post-P64 migration
Admin Head-of-PG
OrphanTasksPage
Tasks with no owner. Off-nav.
Management
Fold into Settings
Folds into Settings:Orphans.
  • Admin: Find orphaned tasks
  • Head-of-PG: Spot games with no auditor
Admin Head-of-PG
GamesPage
Game list. Off-nav.
Operational Management
Split
CRUD half goes to Settings:Games. Observe-as-PG half goes to PG Operational per-game tracking.
  • PG: Browse my games + their status
  • Admin: Create / edit / disable a game
PG Admin
ExecutionMonitoring
Agent execution monitoring. Off-nav.
Observability
Fold into M3
  • (TBD)
TBD
ConversationalInsights
Conversational insights view. Off-nav.
TBD
TBD
Either folds into M3 or kill. Verify with Sergey.
  • (TBD)
TBD
Chat
Chat-style interface. Off-nav.
TBD
TBD
Either kill or move to McSett surface. Verify.
  • (TBD)
TBD
TrackingTable / TaskTrackingPage
Task tracking. Off-nav.
Operational Observability
Fold
Fold into Weekly + Settings:Build-log.
  • TL: Track tasks across stages
TL
DeliveryReportPage
Delivery report. Off-nav.
Operational
Fold
Becomes Delivery Report drill-down inside PG Operational + Game Page.
  • PG: Generate per-delivery report
  • Yuval R2: Client report generation
PG Apollo Lead

Bucket A, Operational

#bucket-a

Weekly

Keep

TL primary. V2 MVP, ships M1 (2026-05-11). Mon-Fri kanban grid, auditors as rows, task cards in cells. Already locked.

Absorbs from V1: Revisions backlog (kills the Ideation switch), per-auditor Monday-morning open-tasks (Tatiana P63), Status / Dashboard quota progress, Builds chip per playable, and Carryover from prior weeks.

PG Operational

New, M2

PG primary. Replaces the current Delivery page. Day-to-day work hub.

Two design directions in active design: queue-first (v1) vs game-first (v2). Decided after PG outreach interview. Absorbs Game Page PG-context, customer interaction log, and Send-with-brief swap (P44).

Ideation fragments

Relocated

The Ideation page itself drops. Two operational handles survive and move to PG Operational header actions: Request Ideas (PG asks for more ideas trigger) and Add Prompt (PG adds new prompt to a game's program).

The filterable ideas-flow numbers-view goes to Bucket C observability instead.

Bucket B, Management and Settings

#bucket-b

In scope this round. NEW, large, M2 ships. Back-office hub for entities, allocations, and configuration. Rare access, lots of stuff, all in one place.

Allocations

New

Four matrix views: Auditor↔TL, PG↔TL, Auditor↔PG, and Game↔TL (per Tatiana P64, migrating off auditor↔game). Inline reassign.

Entities CRUD

New

Create, edit, and delete TLs, PGs, Auditors, and Games. Per-entity fields: name, email, Slack handle, role, status, start date.

Games CRUD

From V1 GamesPage

Per-game fields: tier, core status, monthly quota, TL owner, customer Slack channel, brief link, default effort estimates per task type, is-disabled toggle. PG observe-of-games stays in PG Operational; CRUD lives here.

Builds activity log

From V1 Builds

Active builds (real-time queue) + filterable history (game, TL, auditor, status, type) + stuck-builds (queued or building beyond threshold) + capacity vs demand strip. Per-row click opens Build Detail drawer.

Orphans

From V1 OrphanTasks

Games with no TL owner, auditors with no PG assignment, tasks with no owner, playables with no brief link, games with no monthly quota. Click any row to open the relevant entity edit form.

Schedules

New

Per-auditor vacation calendar, sick days (logged retroactively), recurring availability. Currently nowhere in V1 (P33). Used by Weekly to compute capacity overrides automatically.

Capacity defaults

New

Per-task-type effort defaults: Revision (R), Version (V), New Concept (NC). Each gets default hours. Per-playable override on the playable record. Feeds Weekly red-line and TL monitoring dashboard.

Imports / Migration

From V1 Ideation

Ideation Import tab demoted here (P19). Bulk-import games, playables, briefs from CSV or external systems.

Feedback archive

From V1 Feedback

"What feedback did I give?" filterable list. Drill into individual entry shows full text + screenshot + thread. Other candidate homes (sidebar icon, M3 observability tab) still open.

Admin preferences

From V1 Admin panel

System-level settings for admin authority. Lists/tables here are inconsistent across V1 and drive the DS list/table primitive work (AB-058).

Agents (future)

Future

Reserved slot for McSett (autonomous fix-bot) configuration, McMortar, and automation rules. Lands after M3 observability spec.

Bucket C, Observability

Deferred to M3 #bucket-c

Not for now. Will become its own venture later.

Where status gets watched. Pages folding here: Status, Dashboard, Auditors, Revisions-analytics, Scorecard (kept-as-is), Builds-history, Feedback-analytics, Ideation numbers-view, McSett, plus the off-nav ExecutionMonitoring. Per-role consolidation surfaces TBD.

For Sunday: just note the slot exists, don't dive. Will become its own venture under M3 with a dedicated spec.

PG Jobs to Be Done

#pg-jobs

Every PG job we have on file, with proposed solutions for the PG Operational page. Click to expand.

How we measure success

Each job below carries an opportunity (what business outcome the solution unlocks) and a measure (how we test if it landed).

Method shared across all jobs , (1) in-product sensors capture frontend events, (2) PG outreach interviews validate qualitatively, (3) pre-launch baseline vs week-4 trial reading. Business goals reference vault/00-Success-Criteria.md (BG1-BG6).

01 Review my queue (filter to my games only) Unmet

No "filter by my games" exists in V1. PG sees every game's queue.

Proposed solutions
  • v1 queue-first Default landing IS the filtered queue across the PG's games only.
  • v2 game-first Game list shows only PG's games; drill in for queue.
Problems and pain points
  • No filter by my assigned games for PGs.Problem catalogue
  • PG assignment view hard to scan with many people.Problem catalogue
  • Game page lacks clear job-to-be-done.Problem catalogue
  • Revision count mismatch across 4 pages: 34 / 104 / 181 / 244.Problem catalogue, high
  • ~140 ghost items from Monday import polluting the queue.Problem catalogue, high
  • "Cosmo doesn't serve their workflow well. Lowest current adoption of any role."PGs.md L24-25
  • "PGs should be able to filter the assignment view by 'my games' only , currently impossible."PGs.md L33
Opportunity BG4 PGs scan their work in under 10 seconds instead of mentally filtering across all games. Foundation for every other PG action.
Measure Time-to-first-action on landing page (sensor); interview, "first thing you do after opening PG Op?"
02 Send a playable to client Served-ish

Works in V1 Delivery, but SLA visibility is buried per-row.

Proposed solutions
  • both Ready-to-Send tab is the front door, with aging chip + SLA banner.
  • v2 Per-game card surfaces "ready to send" count + ages.
  • action Bulk-send when batched.
Problems and pain points
  • "undefined" appearing in delivery item titles , erodes trust.Problem catalogue (data quality)
  • Ready-to-send items aging 18-22 days idle in the queue.Problem catalogue, high
  • Apollo prompt stays attached when sending a version , PG can't tell which version they're looking at.Problem catalogue, high
  • Delivery tracking structurally incomplete. Clients use Slack/Jira/Trello/Drive but only Slack is integrated; PGs manually bridge.Problem catalogue, high (eng-acknowledged)
  • "Red-alert anything over 2 days."Einat R2 on aging threshold
  • "Replace the Apollo prompt with the short delivery brief on send."Yuval R2 / 2026-04-19-R2 L34
Opportunity BG1BG4 Cut Ready-to-Send aging from 18-22 days to under 2 days (Einat threshold). Closes the trust loop with clients.
Measure Median age of Ready-to-Send when sent (sensor); count of items aged over 2 days week over week.
03 Track quota commitments per game (monthly + weekly) Partial

Scorecard approaches it; metrics unclear; PGs still spreadsheet on the side.

Proposed solutions
  • v2 Per-game card surfaces tier · monthly quota · weekly commit · delivered-vs-quota up front.
  • v1 Right-rail per-game pin shows quota when card selected.
Problems and pain points
  • "In Pipeline: 121.6% of quota" , meaning unclear; PGs don't know what the number says.Problem catalogue, medium
  • Scorecard metrics are unclear , formulas not understood.Problem catalogue, high
  • Revision count mismatch across 4 pages: 34 / 104 / 181 / 244.Problem catalogue, high
  • Revision count discrepancy: 181 vs 244 across views.Problem catalogue, high
  • All "Not Started" cards show identical quota , likely broken data.Problem catalogue
  • Unreliable Status Dashboard , quotas wrong, makes all metrics meaningless.Problem catalogue, critical
  • "Need to track quota commitments, monthly spending, open revisions per game."PGs.md L29
  • "Cosmo weekly quota is currently manual , should derive from monthly quota / 4."PGs.md L43
Opportunity BG3BG4 PG hits weekly target 80%+ instead of guessing; quota-vs-actual is visible per game at a glance.
Measure % of games hitting weekly commitment in week 4 vs pre-launch baseline; interview, "do you trust the quota number now?"
04 Prioritize revisions over new work Unmet

Rule exists ("revisions first") but UI doesn't reflect it.

Proposed solutions
  • both Need-Action tab ranks revisions first by default, with rule visible.
  • v2 Per-game Need-Action sub-tab carries same ordering.
Problems and pain points
  • Revisions prioritized over new work but UI doesn't reflect this rule.Problem catalogue
  • Revisions must be allocated from Ideation, not Weekly , forces constant context-switching between two pages.Problem catalogue, high
  • Revisions turn green after allocation but remain in the pool , unlike tasks which disappear. TLs must mentally subtract.Problem catalogue, medium
  • Extra "press to board" step required before sending a revision , unnecessary intermediate action.Problem catalogue, medium
  • Revision status unreliable , shows "open" even when completed (e.g. 20 days open, done same day).Problem catalogue, critical
  • Tier / difficulty / deadline fields on revisions are invisible to auditors in Apollo , makes the fields useless.Problem catalogue, high
  • "Revision dashboard mixes already-seen-by-partners vs still-in-flight, and customer vs internal revisions, with no distinction."2026-04-28-Yuval-Harel-PG-walkthrough L23-26
  • "Rejections , unclear whether a rejection counts as a new delivery or a revision, which corrupts tracking."Yuval-Harel L26
Opportunity BG1 Clients see revisions resolved within one cycle instead of accumulating. Direct trust signal.
Measure Median revision time-to-resolution; revision-then-new-work order count per PG per week.
05 Surface client feedback patterns Partial

Revisions spec aggregates; live pattern view absent.

Proposed solutions
  • both Customer interaction log per game (port from McSettActivityTimeline).
  • both Filter-to-ideas-flow header action surfaces patterns + counts.
Problems and pain points
  • Missing revision analytics , no aggregation, no patterns surface.Problem catalogue, high
  • Internal revisions (TL/PG rejections) not tracked separately from client revisions , counts inflated, learning signal lost.Problem catalogue, critical (validity questioned)
  • No "feedback" status to distinguish "waiting for feedback" vs "after feedback" , two states needed.Problem catalogue, high
  • "Quiet churn risk , not all clients give active feedback , some quietly churn."R2-stakeholder-summary L67-68 (Fishi)
  • "It's a crapshoot , clients don't understand the reasoning. PGs can't show the why behind delivered work."Harel-Pick-PG-Head.md L49 / R2 L88
  • "First shot is bad , high variance in asset generation; PGs eat the rework."Harel-Pick-PG-Head.md L48 / R2 L84-85
Opportunity BG1BG4 Spot churn signals + repeat-revision patterns before they harden into churn risk.
Measure Count of feedback-pattern insights surfaced per PG per month; interview, "could you spot quiet churn this week?"
06 Track all prompts I've sent to Apollo (queryable history) Unmet

No prompt history surface today.

Proposed solutions
  • both Delivery Report drill-down with full prompt history per game.
  • v2 Per-game card "prompts in flight" count + drill-in.
Problems and pain points
  • No prompt utilization tracking.Problem catalogue
  • No deep-link from Cosmo to its Apollo task , must manually copy task ID, open Apollo, filter and search to verify status.Problem catalogue, high
  • Apollo silently cancels or fails tasks (not auditor actions) with no notification , prompts disappear with no explanation.Problem catalogue, high
  • "Track prompts sent to Apollo as a queryable history."PGs.md L56
  • "Demand signal broken , no way to verify if Request Ideas actually triggered."PGs.md L48
Opportunity BG4 PG accountability over their own creative output. Removes "what did I send?" question.
Measure Prompt-history surface accesses per week; interview, "find your prompt from 2 weeks ago in under 30 seconds."
07 Skip pre-core games in the queue Unmet

Perpetual Not-Started noise. Depends on P68 (cores in Cosmo, eng prereq).

Proposed solutions
  • v2 Game-list filter on core status; pre-core hidden by default.
  • v1 Queue auto-filters out pre-core when toggle on.
Problems and pain points
  • Cosmo has no awareness of Cores , pre-core games appear identical to ready-to-schedule quota across all views. Engineering-acknowledged limitation.Problem catalogue, high
  • No tracking for games that haven't completed first core.Problem catalogue
  • New games without approved cores appear on the Weekly board identically to regular unallocated quota.Problem catalogue, medium
  • "Pre-core games shouldn't show until core is approved."PGs.md L44
  • "TLs must message Anat each time to check whether core is approved before assigning work."P65 source line
  • "Engineering has explicitly listed this as a known limitation."Eng wiki 2026-04-21
Opportunity BG4 Signal-noise reduction in PG queue. Eng prereq P68 (Cores in Cosmo).
Measure Pre-core games hidden by default count; zero "perpetual Not Started" complaints in interview.
08 Auto-derive weekly quota from monthly (monthly / 4) Unmet

Fully manual today: PG hits +, picks a game, sets a count.

Proposed solutions
  • settings Settings 3.3 holds derivation rule (monthly / 4).
  • v2 Per-game card shows derived weekly commit; quick-edit inline override.
Problems and pain points
  • "In Pipeline 121.6% of quota" , meaning unclear; suggests no derivation rule wired to monthly target.Problem catalogue, medium
  • All "Not Started" cards show identical quota , data layer not driving derivation.Problem catalogue
  • "PG hits +, picks a game, sets a weekly count. Plan-for-future-weeks works in theory but rarely used."2026-04-28-Omer-Peleg L36
  • "Auto-derive weekly quota from monthly , should be a Settings derivation rule."PGs.md L43 / R2 L119 / Yuval-Harel L42
Opportunity BG3BG4 Kill manual weekly-quota entry. PG never sets weekly count by hand again.
Measure % games with derived quota active; manual-entry events per week (target zero).
09 Manage process across the studio relationship (land-and-expand) Unmet

Harel: "PG job is land-and-expand. Delivery itself is the most minor thing." No Cosmo surface treats studio as the unit.

Proposed solutions
  • v2 Game cards group by studio/customer; studio rollup card per client.
  • v2 Studio collapse shows total games + open revisions + recent customer activity.
Problems and pain points
  • "PG job is land-and-expand. Delivery itself is the most minor thing; most of the job is process management."Harel, 2026-04-28-Yuval-Harel-PG-walkthrough L14-15
  • "Track what's happening with a specific client across all their games , huge pain."R2-stakeholder-summary L38
Opportunity BG2BG3 PG retains studio relationship long-term, lands-and-expands. Direct revenue lever.
Measure Per-studio rollup card adoption; interview, "do you know what's happening across all of your client's games?"
10 Add a new prompt to a game's program Partial

Exists in V1 Ideation; discoverability and use unclear.

Proposed solutions
  • v1 Header action "Add Prompt" opens modal with game + brief + body fields.
  • v2 Per-game header action scoped to the game in context (cleaner mental model).
Problems and pain points
  • No way to request prompts from PGs , Request Ideas button exists but discoverability and use unclear.Problem catalogue (partially wrong)
  • No indication of how many prompts are missing relative to plan in Ideation , PGs have no way to know if behind on prompt creation.Problem catalogue, medium
  • No automated notification to PGs when ideation is missing for upcoming-week games , gap discovered mid-week by auditors with no time to fix.Problem catalogue, high
  • Tasks without ideation indistinguishable from tasks with ideation in assignment UI , auditors find no prompt attached.Problem catalogue, high
  • Stale prompts accumulate in Atlas without pruning , 2-3 weeks old mixed with fresh.Problem catalogue, medium
  • Import tab too prominent in Ideation (only used for Mande migration).Problem catalogue
  • "Ideas from 2-3 weeks ago are not relevant."Barzel R2 / R2-stakeholder-summary L98
  • "Demand signal broken , Request Ideas doesn't surface anywhere visible."PGs.md L48
Opportunity BG3 Demand signal closes the loop: Request Ideas → prompts in flight → Apollo throughput.
Measure Missing-prompt count per week (target zero); Add Prompt invocations per game per week.
11 See cross-system status (Apollo + failed items) from a PG cut Unmet

Missing visibility into Apollo statuses + failed items.

Proposed solutions
  • both Aging tab + SLA banner show queue-side health.
  • v2 Per-game card build-status pill.
  • m3 Full cross-system observability lands in Pulse.
Problems and pain points
  • Silent failures on cross-game concept prompts , no notification.Problem catalogue, high
  • External-phase status invisibility , Apollo, Atlas, Slack stages don't surface on the playable in Cosmo. Managers/PGs/TLs go blind mid-lifecycle.Problem catalogue, high
  • Apollo silently cancels or fails tasks (not auditor actions) with no notification.Problem catalogue, high
  • Delivery tracking structurally incomplete. Clients use Slack/Jira/Trello/Drive but only Slack integrated.Problem catalogue, high (eng-acknowledged)
  • No build failure notifications.Problem catalogue
  • Build failures lack ownership context , TLs get no build notifications at all.Problem catalogue
  • "Missing visibility into Apollo statuses + failed items."PGs.md L34
  • "Cross-system invisibility means PGs go blind mid-lifecycle."Yuval framing / Jobs-to-be-Done.md L322-325
Opportunity BG4 PG stops "going blind mid-lifecycle." Reduces Slack-asking-around for status.
Measure Cross-system status surface adoption; interview, "do you still ping Slack to find out what happened?"
12 See ETA / ready-to-track vs actively-worked vs stalled Unmet

Harel: "ETA is invisible. No clean way to see ready vs actively worked vs stalled."

Proposed solutions
  • both New status taxonomy: ready-to-track, in-progress, stalled, waiting-client, delivered.
  • both Per-card status pill with explicit ETA when known.
Problems and pain points
  • Revision status unreliable , shows "open" even when completed (e.g. 20 days "open" while done same day).Problem catalogue, critical
  • External-phase status invisibility , Apollo, Atlas, Slack stages don't surface on the playable in Cosmo.Problem catalogue, high
  • ~140 ghost items from Monday import , poison status reads.Problem catalogue, high
  • No comprehensive task-detail view , list shows only "a version for Merch Cruise" without task name, forcing click-in-click-out.Problem catalogue
  • No deep-link from Cosmo revision to its Apollo task , can't verify status without manual hop.Problem catalogue, high
  • "ETA is invisible. No clean way to see when a playable is ready to track vs actively being worked vs stalled."Harel, 2026-04-28-Yuval-Harel-PG-walkthrough L28
  • "Status view is not intuitive."PGs.md L34
  • "Hard to understand weekly status / what to expect."PGs.md L36
Opportunity BG1BG4 PG can promise the client a date. Trust signal + reduces follow-up churn.
Measure ETA accuracy (set vs actual); count of items with ETA filled per week.
13 Stop manual file renaming and feedback copy-paste Out of IA scope

Operational tax. Eats real PG time. Data plumbing problem, not IA.

Proposed solutions
  • flag Eng/automation backlog. PG Operational can't fix this directly.
  • future Auto-rename on send + structured feedback capture form.
Problems and pain points
  • "Manual file renaming and feedback copy/paste eat real time."Harel, 2026-04-28-Yuval-Harel-PG-walkthrough L29

Not catalogued. Operational tax, IA cannot fix directly. Eng / automation backlog.

Opportunity BG2 PG time recovered from operational tax. Out of IA scope; eng/automation domain.
Measure No measure this venture. Flag for eng intake.
14 Scorecard breakdown per individual PG (own performance) Unmet

Scorecard is portfolio-wide today; no per-PG cut.

Proposed solutions
  • m3 Per-PG Scorecard cut lands in M3 observability.
  • nav PG-Op header link to "My Scorecard" surfaces it.
Problems and pain points
  • Scorecard metrics are unclear , formulas not understood.Problem catalogue, high
  • Revision count mismatch across 4 pages: 34 / 104 / 181 / 244.Problem catalogue, high
  • "In Pipeline 121.6%" , meaning unclear.Problem catalogue, medium
  • Internal revisions (TL/PG rejections) counted twice with client revisions , inflated counts, lost learning signal.Problem catalogue, critical
  • "Monthly scorecard needs per-PG view."R2-stakeholder-summary L41
Opportunity BG1BG4 PG owns their performance, sees what to improve. Self-driven quality.
Measure Per-PG Scorecard view-rate per week; interview, "do you know how you're doing this month?"
15 Create client-facing delivery reports Unmet

No report generation in Cosmo today.

Proposed solutions
  • both Delivery Report drill-down generates a client-shareable summary from delivered items + briefs + customer feedback.
  • v2 Per-game card → Delivery Report shortcut.
Problems and pain points
  • Delivery tracking structurally incomplete , reports would be partial because only Slack is integrated.Problem catalogue, high
  • "Need ability to create reports for clients."R2-stakeholder-summary L37
  • "Ability to create client reports , clients need delivery summaries but no report generation in Cosmo."Yuval-Balak-Apollo-Lead.md L62
  • "Planned quota vs actually executed / sent to client / received a client revision, per game / PG / total."Yuval eng wiki, L77-79
Opportunity BG5 PG ships reports in minutes not hours. Client-trust + sales lever.
Measure Report generation count per week; median time-to-generate.
16 Track what's happening with a specific client across all their games Unmet

Yuval R2: "Huge pain , tracking what's happening with a specific client."

Proposed solutions
  • v2 Game cards group by customer; studio rollup card.
  • v1 Cross-game queue tagged by customer; filter chip.
Problems and pain points
  • "Huge pain: tracking what's happening with a specific client."R2-stakeholder-summary L38
  • "Client-level cross-game roll-up needed but no surface today."Yuval-Balak-Apollo-Lead.md L63
Opportunity BG2BG5 PG owns client-level health. Cross-game patterns surface.
Measure Studio-rollup card adoption; reduction in "huge pain" qualitative score in interviews.
17 See game/library view of only delivered PAs (not in-progress) Unmet

Yuval R2: "In game/library view, show only delivered PAs, not in-progress builds."

Proposed solutions
  • v2 Per-game card surfaces delivered count; drill shows delivered-only filter.
  • filter Library toggle: "Delivered only" vs "All in-flight."
Problems and pain points
  • Game page lacks clear job-to-be-done.Problem catalogue
  • No comprehensive task-detail view , forces click-in-click-out.Problem catalogue
  • "In game/library view: show only delivered PAs, not in-progress builds."R2-stakeholder-summary L42
Opportunity BG5 PG can show portfolio of shipped work to client without in-progress noise.
Measure Delivered-only filter usage; interview, "is this client-ready right now?"
18 Reject / promote items in the line (queue prioritization) Partial

Harel: "90% of PG time is shipping; the rest is prioritization, rejecting, promoting." Exists in V1 Delivery, not surfaced as a primary job.

Proposed solutions
  • both Queue is the prioritization surface. Drag-reorder + reject/promote actions on each row.
  • v2 Per-game queue scope (narrower mental load per session).
Problems and pain points
  • No task ordering or prioritization in Apollo , auditors see all tasks scattered and mixed up with no priority signal.Problem catalogue, high
  • "Review" status color doesn't distinguish TL vs PG queue , TLs can't tell which queue holds the task.Problem catalogue
  • Revisions prioritized over new work but UI doesn't reflect.Problem catalogue
  • "90% of PG time in Cosmo is shipping; the rest on prioritization, rejecting, promoting items in the line."Harel, 2026-04-28-Yuval-Harel-PG-walkthrough L18
  • "Solitaire-Cash example: client said 'do revisions only'; PG/TL had to manually decide who works only on revisions , the system didn't help route this."PGs.md L60 / R2 L108
Opportunity BG1 PG enforces priority signal at the system level. Solitaire-Cash-style routing fixed.
Measure Queue reorder events per week; Solitaire-Cash-style routing time-to-resolve.
19 Bring video PA workflow into Cosmo (currently lives on Monday) Out of M2 scope

Harel: "Videos still live on Monday, uncomfortable and complicated." Wishes video into Cosmo.

Proposed solutions
  • roadmap M4+ once playable schema accommodates video PA records + delivery channel.
Problems and pain points
  • "Videos still live on Monday, uncomfortable and complicated. Harel's wish: bring videos into Cosmo too."Harel, 2026-04-28-Yuval-Harel-PG-walkthrough L16-17

Not catalogued. Schema/scope problem; flag for roadmap.

Opportunity BG2 Video workflow consolidates from Monday into Cosmo. Out of M2 scope, roadmap M4+.
Measure No measure this venture. Flag for roadmap intake.