A/E design review — product brief (discovery)

Product context: Estym8 is built from scratch as an AI-first construction preconstruction platform. This brief is a separate job head on the shared plan-reading spine — for design / engineering firms, not GC estimating and not municipal permitting.

Status: Discovery (2026-07-20; refreshed 2026-07-24 with Paul’s extensive notes; naming 2026-08-05).
Positioning: Same plan-reading engine; separate product surface preferred for GTM clarity (Estym8 module remains an option if Paul wants one login). Not a bid estimate; not city permit stamp. Can run city-style pre-checks before submit (dual-sided with municipal ColdEyes / ColdEyes rule packs).
Working brand: PlanM8 (public 2026-08-28 — Estym8 *8 family; host planm8.ai).
Prior names: ClearToIssue (working) → CheckSet (2026-08-05 direction) → PlanM8. AEC “check set” remains the job language in copy (“Run the check set. You still issue.”). Parallel to municipal ColdEyes.
Do not confuse this SKU with Prisma PlanM8* models — those are ColdEyes DB internals.

Related (public): suite map · ColdEyes · AI-first positioning · investor overview.
Repo only: paul-notes-construction-ai-suite-2026-07-24.md.

---

One-liner

Ingest a design set (or revision package) → walk reviewers through a structured QA workflow that flags internal drawing conflicts, completeness gaps vs the firm’s QA checklist, and coordination issues so the A/E can catch problems before they ship to the owner, GC, or AHJ. AI drafts; humans confirm.

---

Job (and what it is not)

Success metric = issues caught pre-issue + review cycle time + trust (false-positive rate), not F4c vs human takeoff.

---

Who buys / who uses

Buyer = design / engineering firm (or multi-office practice). User = internal reviewers and discipline leads.

---

Shared spine vs separate head

Same pattern as municipal:

[ Shared ]                         [ A/E design-review head ]
folder / PDF ingest           →    project / phase / revision intake
classify + sheet intel       →    firm deliverable checklist completeness
legend / schedule / organic   →    evidence for coordination conflicts
concerns / conflict log       →    internal redline / comment packages
                              →    assign, audit, learn from staff pins

Architecture rule: shared packages; separate auth/tenancy/pricing if sold as its own brand — or a tightly scoped Estym8 module if Paul prefers one login for estimating + design QA.

---

v1 capability sketch (aligned with Paul 2026-07-24)

Paul’s design QA today already mirrors the shared spine: full project ingest → municipality/requirements → inconsistencies / code / constructability → human investigation list → track corrections across submissions.

1. Package intake — multi-PDF / folder; revision + design phase labeling (SD → 50/75/90/95 → permit → IFC).
2. Document-control pre-check (MVP priority) — title block, address/parcel, sheet index completeness, missing/duplicate sheets, code edition, seals, broken sheet refs (Paul: common immediate rejection).
3. Internal conflicts — schedule ≠ plan, callout→detail mismatches, mixed revisions, discipline coordination.
4. City-style pre-check (optional) — same rule packs as ColdEyes (“what the city will bounce”) before submit.
5. Tabular comment log — sheet / location / issue / plan ref / code ref / discipline / severity / status (AI pre-populates). Shared evidence chain: lib/suite/decision-trace.ts (where → measured → required → margin → verdict).
6. Reviewer workflow — confirm/dismiss/add; export comment PDF/JSON/spreadsheet.
7. Phase memory — retain prior-phase decisions; flag regressions (e.g. electrical room later violated).
8. Staff-added findings → gated learning — same as municipal.

Defer: full PE stamp replacement, GC estimating, Revit GA (spike first — measured vs required + edition-locked citation; not judged if data missing; not a chatbot HUD), deep civil geometry day one.

---

Discovery questions for Paul — status after 2026-07-24 notes

• Workflow mirrors municipal review + back-check interest — see paul-notes…
• Dual customer: internal QA and pre-submit “city-style” checks
• Document-control / index checks are early wins
• Evidence + code citation required on every flag
• Revit preferred long-term; PDF vector next-best
• Revit spike bar (Paul 2026-09-01) — decision-trace + abstain if data missing; not waitlist copy; do not promise live Revit code compliance. Queue: CTI24 / S8 / PM8-20.
• Buyer packaging: separate brand (PlanM8) vs Estym8 module?
• v1 checklist source: firm standards vs shared municipal packs first?
• Pilot success metric (hours saved, first-pass comment reduction)?
• Integration must-haves (Bluebeam / Autodesk / Procore export)?
• Must-nots (liability language, stamp adjacency)?
• Public cutover: PlanM8 landing /planm8 + host planm8.ai (no longer a ColdEyes alias)

---

Naming (working → rename)

Suite (current brands): Estym8 (company + takeoff) · ColdEyes (cities) · PlanM8 (A/E pre-issue QA) · Bodi · Feasibility

Tagline: “Run the check set. You still issue.” · “Internal design QA on the same plan-reading engine as Estym8.” · “Not an estimate — a pre-issue review assist.”

Cutover: public surface is PlanM8. Do not use ColdEyes for this SKU. Host planm8.ai / www.planm8.ai is on the estym8 Vercel project; Namecheap records match estym8.ai / coldeyes.ai (apex A 216.150.1.1 + 216.150.16.1, www CNAME to the project vercel-dns hostname). checkset.ai remains optional if we ever lead with the CheckSet wordmark.

Portfolio note (investors)

Investor pack includes this SKU as adjacent discovery alongside municipal ColdEyes and Bodi. Raise stays Estym8-anchored. Prefer saying PlanM8 + ColdEyes in new materials.

---

Document history