Estym8 — prioritized backlog (partner view)
Audience: Paul, Jonas, and other design / GC / city partners.
Updated: 2026-09-01.
Share link: estym8.ai/docs/partner-backlog.
Internal source of truth:UNIFIED_BACKLOG.md(Estym8 Now) ·SUITE_BACKLOG.md(full queue per suite product).
Product context: Estym8 is an AI-first construction preconstruction suite — folder-native plan reading on one spine for Estym8, Bodi, ColdEyes, PlanM8, and SiteM8. The lean raise funds suite development; Estym8 is the near-term revenue wedge. Adjacent jobs (diligence, addenda deltas, shop vs issued, discipline packs) are ranked on the expansion board — modes and packs, not extra brands this quarter.
Path to testing (how we sequence)
1. Pipeline solid → organic takeoff accuracy on real jobs (Jonas / GC)
2. Workflow / UI / UX → project → folder → review → export usable for testers
(includes **ROM / WAG** chase-vs-pass screening UX — P1 #5b)
3. Closed beta → expand testing (free / approved access)
4. Post-beta → Stripe charging · RSMeans cost API (when procured)
We will not turn on paid Stripe Checkout as a near-term priority — billing waits until we are out of beta testing. RSMeans / licensed cost data is on the backlog (procurement-gated); it does not block takeoff accuracy or beta.
Municipal (ColdEyes): P1 ACTIVE — municipal design-partner path (elevated 2026-08-04; partner city name private). Beta landing: /coldeyes. Public brief: /docs/municipal-plan-check. Full queues: suite backlog §3.
How to read priorities
| Rank | Meaning |
|---|---|
| P0 | Pipeline solid — takeoff accuracy |
| P1 | Workflow / UI / UX + security → ready for beta testers |
| P2 | Broader validation, corpus, ops polish |
| P3 | Suite products — PlanM8 / Bodi / SiteM8 (full backlog); expansion ranks later |
| P4 | Post-beta / later — Stripe, RSMeans API, optional depth |
Where we stand (anchors)
| Project | Status | What it means for partners |
|---|---|---|
| 7 Brew — Littleton | Done / locked | Signed-off electrical takeoff baseline |
| Dogtopia — Longmont | Open — accuracy | Packaging + power-plan counting restored; next prove closes duplex gap |
| 7 Brew — Castle Pines | Open — accuracy | Specialty devices after Dogtopia |
Human BOMs / takeoffs are for validation only.
P0 — Pipeline solid (Estym8)
Primary focus for Jonas / GC calibration.
| # | Item | Status |
|---|---|---|
| 1 | Dogtopia — next prove | Ready after mechanism fix |
| 2 | Dogtopia — stability ×3 | After #1 looks honest |
| 3 | Castle Pines — accuracy | After Dogtopia #1 |
| 4 | Model fitness checks | Ongoing |
Partner ask (Jonas): Keep feeding human takeoff / BOM packages for compare-after-run.
P1 — Workflow / UI / UX → beta testing
| # | Item | Why it matters |
|---|---|---|
| 5 | Workflow / UI / UX for beta (folder → review → export) | Testers can complete the job without founder babysitting |
| 5b | ROM / WAG UX — chase-vs-pass screening UI + promote to full bid | Estimators’ high-volume filter (~200 opps/mo); engine MVP exists, product workflow still to finish |
| 6 | Security audit | Safe for wider design-partner access |
| 7 | Onboarding / landing quality for testers | Trust and first-run clarity |
| 8 | Expand closed beta (Jonas + lined-up candidates) | Free / approved access — not paid Checkout |
P2 — Broader validation & polish
| # | Item | Partner angle |
|---|---|---|
| 9 | Product walkthroughs | Smoother demos |
| 10 | Folder-pipeline corpus | Regression across project types |
| 11 | Soccer Stadium AC / surveillance over-count | Paul feedback |
| 12 | Estimate → submittal polish (optional) | Cleaner GC-facing packages |
P3 — Suite products (full backlog)
Same plan-reading engine, different jobs. Not city cost estimating. Not auto-approving permits. Not replacing city reviewers — ColdEyes makes staff more efficient and better at their jobs; humans keep the stamp.
UX: Cohesive workflow apps (queues, checklists, comment logs, back-check) with AI assist — not chatbot-first products.
Competition: Strong point tools in municipal / takeoff lanes — but no AI-first suite peer assembling takeoff + city review + design QA + ICT BOD on one engine.
Detail (IDs, Now/Next/Later per product): suite backlog · investor framing: suite value.
P1 — ColdEyes / municipal design partner (active)
| # | Module | Working name | Buyer | Near-term slice |
|---|---|---|---|---|
| 9 | Municipal permit review MVP | ColdEyes → ColdEyes | Design-partner city | Completeness → comment log → audit → deficiency export |
| 9b | Resubmittal back-check | ColdEyes | Design-partner city | Original + comments + revised |
| 14 | Revit / BIM import (HUGE) | ColdEyes (+ shared spine / PlanM8) | Cities (+ A/E later) | Spike after municipal MVP — traceable measured-vs-required, not judged if data missing |
Partner ask (municipal design partner — now): Confirm pilot scope + success metrics; connect Shannon with city architect for golden packages + audit trails; preferred first permit types (often commercial TI). Partner ask (Paul): ClearToIssue / PlanM8 stays P3; help design-partner workflow / checklist shape when asked. Revit spike after municipal MVP — the bar is a traceable model check (measured vs required + citation), not a chatbot HUD. If a parameter is missing, the check is not judged. PDF / folder first. Humans still issue and stamp.
P3 — Other suite (not municipal-blocking)
| # | Module | Working name | Buyer | Near-term slice |
|---|---|---|---|---|
| 15 | Design agency / A/E QA + pre-submit | ClearToIssue → CheckSet (pairs with municipal ColdEyes) | Design / engineering firms | Firm checklist + city-style pre-check |
| 16 | Site feasibility | Feasibility | Architects / developers | Parcel / zoning → setbacks |
| 17 | ICT project brain → BOD | Bodi | ICT / LV | Shared auth + Estym8→Bodi feeds |
P4 — Post-beta / later
| # | Item | Notes |
|---|---|---|
| 17 | Stripe go-live | Only after beta testing; billing code may already exist |
| 18 | RSMeans / licensed cost API | On backlog; integrate when procurement + API terms are done |
| 19+ | Panel/load checks, geometry, e-sign, accounting export | Scheduled when pulled into active queue |
What already works
- Full project folder ingest — classify every file, then takeoff with context
- Electrical device takeoffs — accuracy improving on held-out jobs
- Concerns / conflict flags with human-in-the-loop
- Multi-model pipeline with fitness checks
- Estimates, proposals, estimate-to-submittal packages
- Written product briefs for municipal + A/E suite discovery
- Stripe integration shipped in product code — charging deferred until post-beta
Working agreement
- Pipeline → UX → beta before paid monetization.
- Stripe is post-beta — do not prioritize Checkout go-live during closed testing.
- RSMeans is planned — procurement-gated; does not block organic takeoff.
- Municipal may jump the queue if a design-partner city commits a pilot — otherwise stays discovery.
- Human takeoffs = validation only — never seed pipeline answers from the BOM.
Market signal — do not build (2026-08-07)
Job market is full of “Construction AI Operations Engineer” roles (e.g. drywall/metal-framing subs) whose real job is: stop rekeying OST/Quick Bid → BuildingConnected → Procore → prefab/field via Power Automate / Dataverse / Power BI.
| That hire owns | Estym8 suite response |
|---|---|
| Microsoft low-code glue, Procore/OST APIs, cost-code governance | Out of scope — not a fifth product |
| Bid → buyout → prefab data plumbing | Not ColdEyes / CheckSet / Feasibility / Bodi |
| Optional upstream: sheet-cited takeoff quantities | Estym8 only — deepen finishes/framing verticals when we chase self-perform interiors; export hooks to Procore/OST stay future maybe (partner/integration), not a Power Platform rebuild |
Takeaway: Demand validates the handoff pain, not a new SKU. Stay the AI plan-intelligence node; let platforms + that hire own the glue. See UNIFIED_BACKLOG.md Appendix B (Procore/OST export).
Document history
| Date | Note |
|---|---|
| 2026-07-24 | Initial partner backlog |
| 2026-07-24 | Path to beta; Stripe post-beta; RSMeans named; municipal contingent (design partner) |
| 2026-08-07 | Non-goal note: sub “AI ops” / Procore–OST glue jobs ≠ suite product |
| 2026-09-01 | Paul forwarded AEC Revit × code-check demos; Revit spike bar = trace + abstain, not HUD |