Implement the following plan:

Generic partner roles with index-based content routing

Summary

Replace driver/observer with { "partner": "*" } role config. Add an index prop to GroupRoleContent for routing different content to different partners by member position. Add an overflow prop to GroupActivity controlling how 3rd partners are handled.

Scope: Component changes are general; MDX changes are today's activity only (2026-02-03).

Design

• roles={{ "partner": "*" }} — everyone gets role "partner", no role labels to display
• <GroupRoleContent role="partner" index={0}> — shows to first partner only
• <GroupRoleContent role="partner" index={1}> — shows to second partner only
• 3rd partner overflow controlled by <GroupActivity overflow="random"> (default) or overflow="expand"
  • "random": 3rd member sees one of the existing partners' content (deterministic hash of groupId)
  • "expand": 3rd member keeps index 2, for explicit index={2} bonus content blocks

The existing role assignment algorithm (calculateUserRole/calculateRoleAssignments) needs no changes — with { "partner": "*" } everyone gets role "partner". The index-based filtering is purely a rendering concern in GroupRoleContent.

File changes

1. apps/web/components/mdx/group-activity-context.tsx

Add targetSize and overflow to GroupActivityContextValue:

export interface GroupActivityContextValue {
  // ... existing fields ...
  targetSize: number;                      // NEW
  overflow: "random" | "expand";           // NEW
}

2. apps/web/components/mdx/group-activity.tsx

Add overflow prop to GroupActivityProps and pass through:

interface GroupActivityProps {
  // ... existing ...
  overflow?: "random" | "expand";  // NEW, default "random"
}

Pass overflow={overflow ?? "random"} to GroupActivityClient.

3. apps/web/components/mdx/group-activity-client.tsx

• Accept overflow prop in GroupActivityClientProps
• Add targetSize and overflow to the context value (lines 302-313)

4. apps/web/components/mdx/group-role-content.tsx

Add index prop and position-based matching logic:

interface GroupRoleContentProps {
  children: React.ReactNode;
  role: string;
  index?: number;  // NEW — match by member position within group
}

When index is provided:

1. Calculate member's position in sorted joined members (by roleIndex)
2. If position >= targetSize (overflow member):
   • overflow === "random": effectiveIndex = hash(groupId) % targetSize
   • overflow === "expand": effectiveIndex = position (raw)
3. Match if userRole === role AND effectiveIndex === index

When index is not provided: current behavior (match on role only).

Add a simpleStringHash helper for deterministic overflow remapping:

function simpleStringHash(str: string): number {
  let hash = 0;
  for (let i = 0; i < str.length; i++) {
    hash = ((hash << 5) - hash + str.charCodeAt(i)) | 0;
  }
  return Math.abs(hash);
}

5. content/spring2026/meetings/2026-02-03-assessments-and-agents.mdx

Stage 1 — Review Assessment Design (lines 44-77):

roles={{ driver: 1, observer: "*" }}  →  roles={{ partner: "*" }}

• <GroupRoleContent role="driver"> → <GroupRoleContent role="partner" index={0}>
• <GroupRoleContent role="observer"> → <GroupRoleContent role="partner" index={1}>
• Remove "Driver:" / "Observer:" prefixes from content
• Add brief context about what partner is doing (e.g., "While your partner reads the blog post...")
• <TextSubmission> stays inside index={1} block (the blog reader captures observations)

Stage 4 — Agents Talking to Agents (lines 131-177):

Same pattern: roles={{ partner: "*" }}, replace role props with index props, remove prefixes.

Files NOT modified

• group-activity-context.tsx calculateUserRole/calculateRoleAssignments — no algorithm changes needed; { "partner": "*" } works with existing fill-role logic
• apps/api/src/routes/group-activities.ts — server role calculation unaffected
• group-activity-role-display.tsx — not used in any MDX content
• Other meeting MDX files — scoped to today only

Verification

1. bun run tsc — type check
2. bun run lint — lint check
3. CI=true bun run test:e2e — existing E2E tests pass (group-activity, full-activity-flow)
4. Manual: dev server shows today's activity with no role labels, each partner sees only their own instructions

If you need specific details from before exiting plan mode (like exact code snippets, error messages, or content you generated), read the full transcript at: ~/.claude/projects/-Users-challen-www-usingandunderstanding-ai/ac3e57f2-fad0-4273-a526-5a358b26e14e.jsonl

Do we need to update the e2e test for today's activity?

Ah, I thought we had an e2e test specifically for today's activity.

Having per-activity tests would ease my mind before class starts, yes. We can set these up to only run when requested and not every time. Try to test a complete flow using multiple students, maybe two groups, one pair and one trio, and the instructor controls, and so on. For the assessment part maybe just make sure that link works and the backend responds, but no need to say anything intelligent.