Skip to content

roadmap-10x.md

Latest commit

 

History

History
43 lines (33 loc) · 3.38 KB

roadmap-10x.md

File metadata and controls

43 lines (33 loc) · 3.38 KB
Raw

The 10x Plan

Supersedes docs/roadmap-scale-1m-to-10m.md (parked before the Infinity merge). Status: planned. Recall with "execute the 10x plan" and we start at Multiplier 1.

Where we are after the merge

The app is no longer a single-feature prototype — it's a full window-install operating system: install brain (fit/condition gate, briefing, capture, AI tips/how-to), crew dispatch + My Work, learning flywheel (rollups, learned dispatch, prediction, analytics), training + clearance + education/glossary with spaced repetition, time clock/payroll, job costing/margin + bid calculator, points/gamification, safety/tools/supplies, QC sign-off, and a role model (installer/foreman/admin/Big Boss).

10x is no longer "more modules." It is three multipliers:

flowchart TD
  m1["Multiplier 1: make the intelligence real"] --> outcome["10x"]
  m2["Multiplier 2: trust + reliability floor"] --> outcome
  m3["Multiplier 3: productize + data moat + AI-native"] --> outcome
Loading

Multiplier 1 — Make the intelligence real (finish the flywheel)

Carried from the parked scale roadmap; the merge did the modules, not the smarts.

  • Next-skill recommender: route apprentices to the type they're closest to clearing so training manufactures routable capacity.
  • Continuous re-dispatch + pace-vs-target: rebalance the day live using proven per-installer speed; nudge ahead/behind forecast.
  • Zero-rework upstream: push the fit/condition gate to load-out (block loading a misfit/damaged unit); route rework-prone types to proven-clean installers using QC callback data.
  • Bidding maturity: CAD-driven bid estimator (schedule vs install history), confidence bands from variance, win/loss + margin tracking (job costing is already in).
  • Mine the dead signals: movement dwell time, transcript search across a type, photo_findings defect patterns.

Multiplier 2 — Trust + reliability floor (the gate to scale)

  • AI reliability: an ai_jobs queue + retry/backoff + budget guard + graceful fallback so transcription/tips/how-to never silently die (today they 429). Surface AI health.
  • Tenancy + real RLS: company_id on core tables + role-scoped policies (RLS is wide open today). This is the prerequisite for more than one company.
  • CI + integration tests: GitHub Actions (lint/test/build) + first RPC/offline integration tests.
  • Real data: real ~100-type catalog, real planset parsing + DWG conversion (still stubbed).

Multiplier 3 — The actual 10x

  • Productize to multi-tenant SaaS: same code, per-company isolation, seats/billing — sell to other window & door companies. 10x of value created, not just internal efficiency.
  • Data network effect: anonymized cross-company benchmarks ("industry par time for this type is 42m; your crew runs 55m") — a moat no single shop can build.
  • AI-native install assistant: real-time vision QA (check the shim gap / flashing from the photo before submit), predictive defect prevention, and ask-in-the-ear Q&A on the company brain.

Sequencing

  1. Multiplier 1 (finish intelligence) + the AI-reliability + CI pieces of Multiplier 2 — makes the single-company tool smart and bulletproof.
  2. Multi-job command board + bidding maturity — run 10x volume in your own business.
  3. Tenancy -> SaaS -> data network effect -> AI-native — 10x the software itself.

All work continues on cursor/window-ops-100x-upgrade-014b; production items are written to be promoted, not thrown away.