Advisor Assessment — Mario Software (2D/3D Modelling Tools + Reporting)
Opportunity: Help Mario build AI-assisted 3D modelling from 2D drawings plus the associated reporting system, per his two framework documents. Stage at writing: Concept — blocked on inputs (docs not yet readable; deal structure not agreed) · Assessor: ZCode (acting investment advisor) · Date: 2026-09-12 · Updated: 2026-09-15 — docs obtained & read, re-assessment §8 appended Review cadence: fortnightly portfolio flip-through; re-assess fully when the two documents arrive.
1. What we know (and what we don't)
Known:
- Mario has produced two substantial framework documents (titles, sizes ~11–12 KB, last edited 2026-09-11):
- AI-Assisted 3D Modeling from 2D Drawings — Development Framework & Process Map
- Comprehensive Process Mapping Guide — Methodology and Practical Framework
- The ask: you contribute to building the 2D/3D modelling tools and the reporting system built on that methodology.
- Deal structure is undecided — this is currently the biggest open variable.
Not known (must be filled from the docs + one conversation with Mario):
- Target user, domain (AEC? manufacturing? infrastructure?) and drawing types.
- Whether the pipeline is fully automatic (very hard) or AI-assisted with human review (tractable).
- Mario's role, funding, time commitment, and what he expects from you.
- IP ownership and what happens if either party walks away.
⚠️ This assessment is intentionally provisional. Re-score every section
after the documents land in
04_Inputs/(ACT-001) and the scope conversationhappens (ACT-002).
2. The thesis options — paid work vs partnership
Option A — Paid development (you build, Mario pays)
| Dimension | View |
|---|---|
| Revenue | Immediate, per milestone; priced like senior AI/automation development |
| Upside | None beyond reputation and reusable components |
| Risk | Scope creep on technically hard 2D→3D work; payment risk if Mario is bootstrapping |
| Fit | Choose if Mario can fund realistic rates and you need cash flow now |
Pricing discipline: 2D→3D reconstruction is genuinely hard ML/graphics work. Quote per-milestone against a written MVP slice, never time-and-materials against an open-ended framework. Front-load a paid feasibility spike before any commitment to outcomes.
Option B — Partnership (you invest time for equity / revenue share)
| Dimension | View |
|---|---|
| Revenue | Deferred and uncertain; only worth it if the product thesis is strong |
| Upside | Uncapped if the tool finds a real market |
| Risk | Unvalidated market, funding unknown, IP/decision-rights disputes, relationship strain |
| Fit | Only with: written IP terms, vesting for time contributions, defined exit, and a narrow validated MVP |
Partnership hygiene (non-negotiable if Option B): vest your equity over delivered milestones; define who owns the code you write; agree what happens on deadlock; cap your unpaid hours until the next gate.
My advisor lean (current evidence)
Neither yet. The correct move is a time-boxed, paid feasibility spike (5–20 h) after the docs arrive: read both documents, pick one narrow drawing-type pipeline (e.g. one drawing family → parametric 3D → auto-report), and produce a technical approach + effort estimate. That spike is sellable as paid work regardless of which structure follows, and it generates the evidence to price/structure the real engagement.
3. Investment & capital at risk
| Item | Estimate (current, provisional) |
|---|---|
| Cash required | ~$0 (code on your machine) unless hosting/models are needed |
| Founder hours before structure must be agreed | Hard cap: 20 h (disposable exploration only) |
| Capital at risk | Your time — the same hours the other three opportunities want |
| Time-box | Docs + scope conversation within 3 weeks, or park |
4. Stage gates
| Gate | Exit criteria | Kill / park trigger |
|---|---|---|
| Concept → Validation | Both docs in 04_Inputs/ and read; scope conversation held; deal structure agreed in writing; MVP slice defined (one drawing type, one report) | No docs/structure agreement by 2026-10-03 → park |
| Validation → Build | Paid feasibility spike delivered; technical approach reviewed; effort estimate within Mario's funding | Feasibility shows the auto-3D goal needs research-scale effort → reduce scope or walk |
| Build → Market | Working MVP slice used on real drawings by someone other than Mario | Nobody real uses it after 2 demos → reassess with Mario |
5. Top risks (full register: 03_Registers/RSK.csv)
- Technical feasibility overpromised (HIGH) — full automatic 2D→3D is research-grade; mitigate with narrow slice, parametric/rule-based core, AI-assist where reliable.
- Deal structure undefined (HIGH) — every hour before agreement is a gift; hard 20 h cap.
- IP ownership ambiguity (HIGH) — you may bring substantial code; ownership must be written before you build.
- Market unknown (MEDIUM) — established BIM/CAD vendors occupy adjacent space; the docs should say who exactly buys this.
- Key-person dependency on Mario (MEDIUM) — his domain knowledge is the moat and the single point of failure.
6. Concept-stage KPIs
| KPI | Target |
|---|---|
| Documents obtained and ingested | 2 / 2 by 2026-09-26 |
| Deal structure agreed in writing | by 2026-10-03 |
| MVP slice defined (drawing type + report) | by 2026-10-03 |
| Unstructured hours burned | ≤ 20 |
| Feasibility spike sold (paid) | 1 by end of October (moves stage to Validation) |
7. Immediate actions
See 03_Registers/ACT.csv — export the two Google Docs into 04_Inputs/ (links in 04_Inputs/GDOC_LINKS.md), hold the 60-minute scope conversation, then decide the structure. Everything else waits on those three moves.
8. Re-assessment — 2026-09-15 (documents landed)
Both documents were exported to 04_Inputs/ on 2026-09-15 (ACT-001, 11 days ahead of the 09-26 due date) and read in full. What they resolve, what they confirm, and what they still leave open:
Resolved unknowns (were provisional in §1):
- Domain & users: AEC — contractors, estimators and small-to-midsize
builders processing large volumes of 2D DWG/DXF/PDF/raster drawings.
- Pipeline ambition: fully automatic ("autonomously reconstruct") with a
human-in-the-loop review studio as one layer — the research-grade end of the spectrum, not AI-assisted drafting.
- Reporting scope: automated BoQ spreadsheets, clash detection during
estimating, IFC 4.3 / Revit / OBJ export.
- Intended roles: Mario = Product Manager; you = Technical Lead,
accountable (A) for the AI pipeline, geometric engines and IFC mapping per his RACI matrix.
Confirmed risks (sharper than at writing):
- Risk #1 (overpromised feasibility) is now documented, not suspected: the
six-layer pipeline (CV symbol detection, OCR, GNN constraint solving, parametric BIM generation) and every KPI claim (≥98% geometric accuracy, mIoU ≥92%, <45 s/sheet, 85% effort reduction) come with no prototype, no benchmark, no data. Treat all numbers as hypotheses for the slice to test — never as a spec.
- The RACI assumes a team of 5–6 (AI/CV engineer, full-stack dev, BIM
specialist, beta testers) that does not exist. The gap between the org chart and the actual partnership (2 people, one capped at 20 h) is the single most important scope-conversation topic.
- Zero demand evidence inside either document. No customer, no interview,
no market sizing — the AEC pain is real in general terms, but nothing in these docs validates that anyone will pay.
- Doc 1 (process-mapping guide) is generic methodology with a worked
vacation-planning example: it evidences structured thinking, not proprietary AEC insight or customer access.
Still open (docs silent): funding, Mario's time commitment, IP ownership, walk-away terms — exactly the ACT-002/ACT-003 material.
What got easier: Mario's own roadmap names the PoC — "2D single storey floor plan to basic 3D reconstruction." That is the MVP slice §2 wanted, pre-agreed by the author of the vision. ACT-004 becomes ratification, not invention.
Advisor verdict (unchanged, now evidence-backed): paid feasibility spike or nothing. The spike targets his own PoC — single-storey floor plan → basic 3D → BoQ extract, 5–20 h, priced per §2's discipline. If Mario cannot fund that spike, that is itself the market signal: park on the 2026-10-03 trigger.