Future scope: Shared student memory, today cards, and a card-based site

Status: FUTURE SCOPE — not scheduled, nothing built. Added by Ravneet, 4 Oct 2026. Related: FUTURE-SCOPE-PILLAR-PAGES.md, ARCHITECTURE.md (bot isolation, opacity).

1. One shared memory per student

When a student logs in, all six bots read the same student profile: name, category, exam, centre, mock scores (Rocky/TCY), classes attended, books, coins, deadlines. No bot asks the student for something Cetking already knows. Each bot uses the same memory for its own job:

  • Rocky: the next tests to take to improve.
  • Guruji: the videos and classes that fix weak areas.
  • Sutra: what to read next.
  • Purva: the student's colleges (see section 3).
  • Arena: the student's own target and challenge.
  • Oracle: sees everything, decides what matters today.

Private fields (category, scores, phone) are seen only by the student, staff and Ravneet; never shown to other students, never on public pages, never given to an AI together with other students' data (opacity rule).

2. Oracle's "today" cards (calm, not crowded)

  • On login, Oracle shows at most 2–3 cards for today, e.g. "Your mock is due tomorrow" (Rocky), "This 12-minute video fixes your weak area" (Guruji).
  • Other bots wait quietly: a small dot on their icon in the six-bot tray. Tapping a bot shows its own card. No bot pushes itself.
  • One card = one action ("Take this test", "Watch this video", "See your 10 colleges"). No paragraphs.
  • Priority rules set by Ravneet (like playbooks), applied by code: exam close → Rocky first; missed a class → Guruji first; admission deadline near → Purva first.
  • Each bot builds its own card (bot isolation): changing Guruji's card can't break Rocky's.

3. Purva's college stack

  • Purva already knows the student's category, exam and latest mock score, so she never asks. She shows a stack of ~10 college cards with only the student's category cutoff, marked Safe / Target / Dream against their own score.
  • Each card opens the full public college page, jumping to their category section (e.g. /colleges/jbims#obc-category).
  • End of stack: "Want to see more colleges?" to widen the list or change filters.
  • No mocks yet → top colleges for their category, with "your list gets sharper after your first mock".
  • Public college pages stay separate and show every category (for Google).

4. Whole site as cards (long-term direction)

Ravneet's vision: every piece of content on all Cetking sites becomes a card — a college, a video, a test, a book, a class, a blog post — that can be stacked and filtered like cells in a spreadsheet, while keeping SEO.

Is this realistic? Yes. This is how large sites already work (property, travel, shopping and college-listing sites): each item is a card, lists are filtered views, and every item still has its own page that Google indexes.

How to do it without it becoming fragile:

  • Every card is backed by a real page with its own URL. The card is a preview; the page is what Google ranks. Cards never replace pages.
  • Cards are generated from database rows, not hand-made HTML. One card design per type (college, video, test, book, class, post) — a handful of designs, not hundreds.
  • Each card type belongs to one bot (college → Purva, video → Guruji, test → Rocky, book → Sutra, challenge → Arena, story → Oracle), following bot isolation.
  • Filters are just views. A few important filter combinations get their own clean, indexable URLs (e.g. /colleges/mumbai, /videos/quant); all other filter combinations are for students only and marked noindex, so Google never sees thousands of near-duplicate pages.
  • Personal views (logged-in) are never indexed. Public pages show the same data to everyone.
  • One shared card frame (size, spacing, tap behaviour) in the shared kit, so all cards feel the same; contents differ per type.

What would make it break (avoid): hand-coded one-off cards; filter pages indexed without limits; personal data mixed into public pages; one giant component that every bot edits.

Open questions (decide when scheduled)

  • First card types to build (likely college + video + test).
  • Exact Oracle priority rules (Ravneet's playbook).
  • Which filter pages deserve their own public URL.

Source: GitHub cetking-one/docs/FUTURE-SCOPE-CARDS.md. Edit the file there; this page updates on the next release.