Repository navigation
README
github-actions[bot] edited this page Jul 20, 2026
·
1 revision
Single page of truth for where the company is. Updated weekly (npm run report:weekly, ideally Friday). Honesty rule is constitutional: every number here is derived from real data or real task state — no aspirational percentages.
- Production: https://classmode.inmodel.in (Azure Container Apps; health check = post-deploy gate)
-
Plan of record: the Sharpened Company 90-day plan (
~/.gstack/projects/NitishKumar-ai-Class-mode/— design doc + CEO plan, /autoplan-cleared 2026-07-15) - Deferred work: TODOS.md · Weekly reports: weekly/ · Metric snapshots: metrics/ · Programs pipeline: PROGRAMS.md
| # | Objective | Deadline | Status |
|---|---|---|---|
| 1 | WTP close attempt at pilot school (audit → term report + paid-pilot offer → ask + who-signs) | first attempt this week; hard fail line day 30 (escalation, not freeze) | 🔴 not started |
| 1b | Same offer at 3–5 additional schools (parallel demand test) | weeks 2–6 | 🔴 not started |
| 2 | Wedge workflow = school's primary record, 30 consecutive school days unprompted | primary-record by day ~40; streak started by day 55 | ⚪ awaiting wedge audit |
| 3 | Dependence probe live: metrics weekly, definitions fixed, baseline captured | preconditions (usage events wired + Gemini billing) by day 14 | 🟡 script shipped; view-events + billing pending |
| 4 | Learning-record schema | GATED on graduation trigger (payment or signed commitment) | ⏸ gated |
| Workstream | Priority | Done | Status / next action |
|---|---|---|---|
| Honesty fixes (fake WhatsApp toast; NULL school_code on admin writes) | P1 | 0/2 | Fix chips spawned — do first, they sit on the measured screen |
| Paid-pilot offer page + buyer-mapping & price-anchor questions | P1 | 0% | Draft before the school visit |
metrics:weekly engine + snapshots |
P1 | 80% | Script + snapshots shipped; wire attendance_view/report_view events |
| Term report (full design+eng spec: hierarchy, thin-data denominators, A4 grayscale, bilingual, aggregate-only, fail-closed queries, hostile tests) | P1 | 0% | Build after wedge audit confirms data shape |
| Custody baseline (FKs off CASCADE, transactional corrections audit, tested restore, CSV export) | P1 | 0% | Schema ALTER first (idempotent pattern) |
| Constitution + DECISIONS.md (incl. segment paragraph, amendment clause) | P2 | 0% | Written the evening AFTER the WTP conversation — not before |
| Vernacular pass (wire t(), locale, switcher, toasts/errors) — language: TBD (TD1) | P2 | 0% | Blocked on naming the pilot school's language |
| Attendance mobile-first reorder (remembered class, sticky save, 44px targets) | P2 | 0% | Pairs with vernacular pass |
| School #2 playbook + conversations | P2 | 0% | Written within a week of first WTP conversation |
| Offline-first attendance (E9 contract) | GATED | spec 100% / build 0% | Post-graduation-trigger by gate decision UC3 |
See the newest file in metrics/ and the table in the newest weekly report. Definitions and decision thresholds live in scripts/metrics-weekly.ts so they can't drift silently. Known probe history: two AI-generation outages produced zero user complaints (both negative signals — nobody depends on AI features yet).
- Moat = daily habit + earned trust + trustworthy custodianship. Never "our AI is smarter." Data is exportable on demand.
- Models are swappable suppliers behind
server/lib/ai/gateway.ts. - No fabricated data to real accounts, ever — including UI copy that claims automation that isn't running.
- WhatsApp/Meta automated pipe stays paused until WTP validated (manual, honestly-labeled previews allowed).
- Day-30 without a written WTP answer ⇒ build pauses, escalation only (real buyer → school #2).