Eurocomponents — Buildability Assessment

Stage: Proposal sent → Zanoni intro (June 8-9)  |  Updated: 2026-06-08
Proposal (9 Giugno): adapttoai-eurocomponents.pages.dev/pre-proposal-external-v3
UI Mockup: eurocomponents-rfq-mockup.pages.dev/index-live.html  |  GitHub: Adapt-to-AI-B2B/eurocomponents-rfq-mockup

Updated Jun 8 — Delta since Call 3

• Pricing restructured — milestone-based: €1K avvio + €1.75K Phase 1 + €3.9K Phase 2 + €1.3K/mo. Old flat impl fee dropped. Rationale in Section 3.
• Technical clarifications from Calls 4–6 — CLS/Lamierino + Posa required on descrittivo header; flag-to-source linking mandatory; output languages locked to IT/EN/FR/DE; supplier pricing formula confirmed; Swiss withholding tax. See Section 2 addendum.
• Pipeline status — proposal sent to Maurizio. Zanoni intro meeting June 8-9. NDA likely required before Zanoni sees full pricing.

1. TL;DR

EC's workflow today — 4 phases, all manual

What the system does

TODAY (all manual, ~6-12 hr per project, throughput-limited)
─────────────────────────────────────────────────────────────────

  RFQ ──► [BO triage]──►[BO descrittivo]──►[Preventivista]──►[BO+Comm offer]──► Client
         30 min-3 hr     10 min-4 hr          ~3 hr              1-2 hr


v1 (AI replaces phases 1+2)
─────────────────────────────────────────────────────────────────

  RFQ ──► [AI: triage + extract + standardize + translate + apply reg rules]
                               │
                               ▼
                      [Human-in-loop QA UI]        ← back-office confirms,
                      (flags linked to source PDF)   every flag → exact page
                               │
                               ▼
                    Clean descrittivo Excel
                               │
                               ▼ (HubSpot notification → preventivista)
                               │
              [Preventivista split + pricing] ──► [BO+Comm offer] ──► Client
                   unchanged ~3 hr                  unchanged 1-2 hr


v2 (extends the platform)
─────────────────────────────────────────────────────────────────

  + Phase 3: split pricing engine (supplier lookup + formula automation)
  + Article code resolution: capitolato → supplier lists (two-stage lookup)
  + Swiss withholding tax (2.5%) handling
  + Auto-photo pickup from supplier sites (Maurizio's #1 ask)
  + NO/SE input parsing + additional country coverage

v1 outcome

• Phases 1+2: 40 min – 7 hrs today → 6-12 min AI + 30-45 min human QA
• 40% of RFQs currently downgraded to "budget" can now flow as real offers
• ~240 projects/yr × ~4.5 hr avg saved = ~1,080 hrs/yr of back-office time freed
• Revenue upside: faster + more complete offers → higher win-rate on pipeline (EC lost a €6.5M project last cycle, has 4-5 months of empty production)

Research Artifacts

What to look at

Real client files — how to access them

These files are currently on Giuseppe's Mac only. Giuseppe needs to share the folder before you can open them. Ask him to send via iCloud Drive / Google Drive / WeTransfer — the folder is ~/Downloads/Mauri & Giuse/ (~50MB total, 4 complete projects).

Once you have the folder, here is exactly what's inside and what each file is for:

Mauri & Giuse/
│
├── EuroComponents_flussocommerciale_AI.pdf          ← START HERE
│     Maurizio's own 4-phase workflow diagram. The spec for what we're building.
│
├── 025-26 Residenziale Hardess Yard - Brixton (UK)/ ← PRIMARY TEST CASE (read this end-to-end)
│   ├── 00_DISEGNI E SPECIFICHE/
│   │   └── 00_DOCUMENTI CLIENTE/
│   │       ├── ...-34000-Typical Bathroom Pod.pdf         ← CAD drawing, typology 01
│   │       ├── ...-34001-Accessible Shower Pod.pdf        ← CAD drawing, typology accessible
│   │       ├── ...-34002-Adaptable Shower Pod.pdf         ← CAD drawing
│   │       ├── ...-34003-Typical Bathroom Pod Type 2.pdf  ← CAD drawing, typology 02
│   │       └── ...-18000-Accommodation Schedule.pdf       ← SOURCE of typology counts (75/213/32)
│   ├── 01_DESCRITTIVO OFFERTA/
│   │   └── 025-26 Rev 00 Descrittivo.xlsx                ← THE CANONICAL SCHEMA (40 cols × 76 rows)
│   │         This is what we output. Open this first.
│   ├── 02_SPLIT/
│   │   ├── Split 25-26 tip 01.pdf                        ← Cost-build per pod, typology 01 (~€7,953/pod)
│   │   ├── Split 25-26 tip 02.pdf                        ← Same for typology 02
│   │   └── Split 25-26 tip 03.pdf                        ← Same for typology 03
│   └── 03_OFFERTA/
│       ├── 025-26 Rev. 00 Hardess Yard ... Brixton.pdf   ← FINAL OFFER (12pp, what the client sees)
│       ├── 025-26 Calcolo prezzo_08-04-2026.xlsx         ← Sell-price calculator (K=5% applied)
│       └── FILE K 025-26 08-04.xlsx                      ← K-floor worksheet (cost floor, K=0 baseline)
│
├── 060-25 Resid.+Hotel Schwarzwaldtrio - Karlsruhe (D)/  ← HARDEST DOCUMENT (German LV, 144pp)
│   └── 060-25/
│       ├── Documenti cliente/
│       │   └── Capitolato cliente/
│       │       └── 2025-06-18_SWT_ LV Fertignasszellen.pdf  ← THE 144pp GERMAN LV
│       │             pp11-13: DGNB ENV1.2/QNG compliance matrix — 30+ DIN/EN norms.
│       │             This is the worst-case input. Read pp1-15 to understand the challenge.
│       ├── Descrittivo Tiziano/
│       │   └── 060-25 Descrittivo offerta.xlsx             ← Karlsruhe descrittivo (41c × 104r)
│       ├── Split/  (4 split PDFs by typology)
│       └── Offerta/
│           └── EC_Rev00_060-25 ... Angebot FNZ.pdf         ← German final offer
│
├── 068-26/                                                  ← THIRD PROJECT (partial — no offerta)
│   ├── 068-26 Rev 00 Descrittivo.xlsx                      ← Another descrittivo (38c × 77r)
│   └── 26005-CRA-XX-00-DR-A-11100.pdf                     ← Client CAD drawing
│
├── 126-25 Residenziale Lucens (CH)/                        ← SWISS PROJECT (French input)
│   ├── 00_DOCUMENTI CLIENTE/  (3 French PDFs)
│   ├── 01_DESCRITTIVO OFFERTA/126-25 Rev 00 Descrittivo offerta.xlsx
│   ├── 02_SPLIT/  (6 split PDFs)
│   └── 03_OFFERTA/
│       ├── 126-25 REV00 Offre sdb préfa - Resid. Lucens.pdf  ← French final offer
│       └── 126-25 Tabella costi e PV.xlsx                    ← Pricing table (IT internal)
│
└── 127-25 Co-living Crissier (CH)/                         ← SWISS CO-LIVING (French input)
    ├── 00_DISEGNI E SPECIFICHE/
    │   └── VD_ARBORA-B11_CFC 259_SOUMISSION_Prefabricated pods.xlsx  ← Client spec in Excel
    ├── 01_DESCRITTIVO OFFERTA/127-25 REV00 Descrittivo offerta.xlsx
    ├── 02_SPLIT/  (2 split PDFs)
    └── 03_OFFERTA/

Suggested reading order for Riccardo:

1. EuroComponents_flussocommerciale_AI.pdf — understand the 4 phases (5 min)
2. 025-26/01_DESCRITTIVO OFFERTA/025-26 Rev 00 Descrittivo.xlsx — open the canonical schema, understand the 40 columns (10 min)
3. 025-26/02_SPLIT/Split 25-26 tip 01.pdf — see what the preventivista does with it (5 min)
4. 025-26/03_OFFERTA/025-26 Rev. 00 ... Brixton.pdf — see the final output (5 min)
5. 060-25/.../LV Fertignasszellen.pdf pp1-15 — understand the worst-case input document (10 min)

2. Engineering Surface Decomposition

Summary

• Total surfaces: 13
• In v1 scope: 8 (#1-7, #12, #13)
• REUSE: 1 (#9 K markup, trivial — but Phase 1 OUT)
• EXTEND: 6 (inbound, language parsing, domain extraction, pricing split, doc gen, approval UI)
• NEW: 6 (document triage, EC defaults library, format-preserving translation, country-reg engine, photo pickup, HubSpot)

Honest read: this is not a low-reuse engagement. ~46% of surfaces are NEW. We have pattern carry-over from Aronlight (same email-to-structured-output + human-in-loop architecture) but the underlying problem inverts (catalog matching → schema extraction).

Highest-risk surface: Country-regulation rule engine (#7)

Why: Karlsruhe LV alone references 30+ DIN/EN/VDI norms with material-category-specific compliance requirements (DGNB ENV1.2 / QNG, EMICODE EC1, GISCODE classes, SCCPs/MCCPs/LCCPs Chlorparaffine ≤ 0.1%, Formaldehyde ≤ 0.10 ppm, etc.). UK Building Regs Part M is implicit in the drawings. CH SIA + IT UNI not yet represented in samples but exist. If we attempt to "build the regulation knowledge" ourselves we are building an entire compliance product, not a feature.

De-risk path (confirmed Call 3 Q6): make the engine a structured rule importer, NOT a regulation-knowledge product. Maurizio writes the rules + supplies legislation files (committed). System ingests them as a structured table and applies them as filters during extraction. Scope: "rule library editor + per-country profile selector + apply rules during standardization step". ~2 weeks of dev for the constrained version.

What the country-regulation rule engine actually is (concrete)

Skip if you've worked with this kind of system before.

A subsystem that, during the extraction + standardization step (Phase 1), applies country-specific compliance requirements as filters and validators on every material line in the descrittivo, then either prevents non-compliant choices, recommends compliant alternatives, or flags them for human review.

Real examples from the materials in hand:

• Karlsruhe LV pages 11-13: matrix titled Anlage 1 — Materialanforderungen für DGNB ENV1.2/QNG. Per material category: VOC ≤ 10 g/l, EMICODE EC1, GISCODE codes (D1, ZP1, RU 0,5...), SCCPs Chlorparaffine ≤ 0.1%, Formaldehyde ≤ 0.10 ppm, references to DIN 18 195/202/336/352/534 + AgBB.
• Brixton drawings: margin notes such as "Bathroom Component Key to comply with Part M Vol 2" — UK accessibility regs.

At runtime, for a Karlsruhe descrittivo:

"Project is German, DGNB Silver target. Proposed tile is Refin Feel. Does Refin Feel have an EMICODE EC1 declaration on file? Does its adhesive meet GISCODE D1? If yes → cleared. If no → surface a substitution recommendation OR flag for human override."

Three sub-components:

Constrained scope (the version we propose): rule applicator, not regulation expert. Client owns the rules + legislation updates. We own the system that applies them.

This is the difference between 2 weeks of dev (constrained) and 2 months + ongoing (unconstrained). The pre-condition "Maurizio commits to writing rule tables" was load-bearing — now resolved (Call 3 Q6).

3. Costs

Build cost — one-time

Bottoms-up by phase. Riccardo = primary builder. Giuseppe + Raffaello = architecture alignment + milestone review only.

Photo-pickup (v2): ~40 hrs dev + €2,000 tooling if feasibility lands. Not in v1 scope.
Discovery + scoping (already done, not in the above): ~20 hrs partner.

Operating cost — monthly

Once live at 240 RFQs/yr pace (~20/mo):

Technical clarifications from Calls 4-6 (addendum Jun 8)

These requirements were confirmed after the surface decomposition above. They affect Surface 4 (domain extraction), Surface 6 (translation), and Surface 9 (human-in-loop UI) specifically.

Net impact on build estimate: The CLS/Posa fields and flag-to-source linking add ~3-5 days dev to Surface 4 (extraction UI + schema). Output language reduction from 6→4 saves ~3 days. Net: roughly neutral on the €18K build estimate.

Pricing model — milestone-based (sent 9 Giugno 2026)

Note: options A/B/C/D explored during internal scoping (May 2026) are superseded. Maurizio requested zero large upfront impl fee (Call 4); pricing restructured as milestone payments tied to deliverables.

Year 1 total: €1,000 + €1,750 + €3,900 + (5 × €1,300) = ~€13,150
Year 2+ ARR: €1,300 × 12 = €15,600/yr

Margin & breakeven (milestone structure)

• Build cost (Phase 1+2, no photo): ~€18,000
• Year-1 gross margin: €13,150 − €18,000 = −€4,850 (invested in delivery)
• Year-2 net: €15,600 ARR − ~€8,280 ops = +€7,320
• Breakeven: ~Mes 20 from kickoff (subscription recovers build shortfall within 14 months of going live)
• 5-year LTV (80% Y2-5 retention): €13.15K + (€15.6K × 4 × 0.8) = ~€62.9K total

Pricing rationale

Why milestone-based: - Maurizio explicitly pushed back on large upfront impl fee (Call 4). Zero advance payment was his constraint. - Milestone gating de-risks the client — they only pay when the system demonstrably works. - Net effect: AdapttoAI absorbs €18K build cost, recovers it via subscription over ~14 months post-delivery. Acceptable given strategic value (Italian beachhead, construction vertical, Zanoni cash position ~€2M). - Floor check: €15,600/yr ARR well above €10K/yr floor.

4. ICP Fit

Reading b2b/docs/icp.md V3 (Apr 16, 2026). Score each condition 0-3.

Minimum ICP fit threshold per icp.md: 7+ "worth pursuing", 10+ "strong fit". EC scores 9 → strong-fit-adjacent.

Pipeline-state context

• Active engagements: Aronlight in 7-week build window. Lamosa Q1 follow-up open. BCP proposal sent. AutoalDía pending. Capacity is constrained — adding 2nd active build risks Aronlight delivery.
• Strategic value: HIGH
• Italian beachhead (most current AdapttoAI clients are LatAm)
• Maurizio has 2nd external client he'd refer once EC works
• Top-down decision-maker (confirmed Call 2 + 3)
• Vertical = construction prefab is underserved → marquee logo potential
• Reference value: Strong. Eurocomponents has 60,000+ pods delivered across 20+ countries. Marquee customer reference for AI-in-construction-tech.
• Revenue urgency: HIGH (was MEDIUM in v1). Call 3 Q5 surfaced explicit production gap (6.5M booked vs 14M target) and 78 offers already done by mid-May vs ~150/yr historic baseline. They need throughput now, not in Q4.

Decision contributions per references/buildability-rubric.md

• ICP score 9/12 = 75% → contributes to GO
• Strategic value HIGH → adds +2 to net score
• Revenue urgency HIGH → adds +1 to net score (was 0 in v1)
• Capacity stretched → adds −1
• Decision-authority real → adds +1

5. Recommendation

Net score (per buildability-rubric § "GO / PASS / PIVOT decision matrix")

Net score: −2 + 2 + 1 + 2 − 1 + 2 + 1 = +5

Threshold: 0 to +3 → PIVOT. 4+ → GO. EC at +5 → GO (smaller shape).

Auto-PASS red flag check

• ❌ No human signer surfaced — FALSE (Zanoni signs, Maurizio carries operational authority)
• ❌ Compliance / regulatory risk we can't mitigate — RESOLVED Call 3 Q6 (Maurizio writes rules, we apply)
• ❌ Hostile decision-maker — FALSE
• ❌ Compete-on-price with $30/mo SaaS — FALSE
• ❌ Custom output format with no relaxation path > 50% of build effort — partial concern (final offer template rigid) but mitigated by leaving Phase 4 manual in v1

No auto-PASS triggers. Verdict stands.

Verdict: PIVOT → GO (smaller shape)

2026-05-07 verdict was PIVOT. Upgraded to GO (smaller shape) on 2026-05-12 after Call 3 resolved 3 of 4 pre-conditions including the load-bearing regulation engine de-risk. Same scope as the PIVOT recommendation; pre-conditions now mostly checked.

Reasoning (5 bullets, citing earlier sections)

1. Regulation engine de-risked — Section 2 surface #7 was the highest-risk NEW surface. Call 3 Q6 confirmed Opzione A: Maurizio writes the rules in our format. This converts the surface from "build a compliance product" to "ingest a structured rule table" (~2 weeks dev instead of ~2 months). The biggest single risk on the engagement is now contained.
2. Volume + urgency upgrade strengthens the economics — Section 3 originally used 150 RFQ/year baseline. Call 3 Q4+Q5 revealed 240/year pace with 40% being "budget" estimates Maurizio wants to eliminate. Higher real RFQ volume + explicit production gap means win-rate uplift is the real lever; labor-cost is the floor.
3. ICP fit upgrades from 8/12 to 9/12 — Section 4 C3 (high-volume) was scored 1 based on 150/year. With 240/year pace, C3 is closer to 2. Pushes ICP into the 75% threshold.
4. Capacity constraint is real but timeline still works — Aronlight build active. EC kickoff scheduled post-Aronlight Mes 3 (July-Aug 2026). Pre-condition #2 (Aronlight clean) is the only unresolved one — independent and on track.
5. Strategic value confirmed — Italian beachhead, Maurizio's top-down decision authority validated, marquee construction-vertical reference. Milestone pricing (€1K avvio + €1.75K Phase 1 + €3.9K Phase 2 + €1.3K/mo) sent June 9 — economics meet the conservative case acceptability bar at Mes 20 breakeven.

GO — smaller shape that works

Smaller shape: - v1 ONLY = document triage + canonical record + standardization (descrittivo Excel render) + multi-language IT-middle automation + constrained country-regulation rule engine. - v1 OUT: photo-pickup (defer to v2 with separate billing), final offer document generation, pricing split engine end-to-end automation, supplier-quote sub-flow. - Smaller country footprint at launch: UK + DE + IT only. Add CH + Norwegian + Swedish in v2 once UK/DE/IT live. - Tighter regulation engine: ingest Maurizio's rule tables, do NOT independently research norms.

Pricing on the smaller shape (sent 9 Giugno 2026): - Fee di avvio (alla firma): €1,000 - Phase 1 completata: €1,750 - Phase 2 completata: €3,900 - Quota Mensile (da Mese 7): €1,300/mo - Year-1 total: ~€13,150 - Year-2 ARR: €15,600/yr - Breakeven: ~Mes 20

Timeline (updated Jun 8): - 2026-05-12 — DONE: Q&A recap sent to Maurizio. - 2026-05-26 — DONE: internal pre-proposal aligned (Giuseppe + Raffaello). - 2026-06-09 — DONE: proposal "9 Giugno" sent to Maurizio. Zanoni intro meeting June 8-9 (NDA likely required before Zanoni sees pricing). - Kickoff: July 2026 (after Aronlight Mes 3 + Zanoni sign-off). - Phase 1 delivery: ~August 2026 (pipeline live on real RFQ, preventivista accepts descrittivo). - Phase 2 delivery: ~September-October 2026 (split pricing + HubSpot wired, handover). - Quota Mensile starts: Mese 7 (~January 2027).

What's cut from original ask: - Auto-photo pickup (Maurizio's #1 emotional ask) — deferred to v2, billed separately, framed as feasibility prototype during M3 delivery. - End-to-end automation (Maurizio knows we're not doing this — explicitly de-prioritized in Call 1). - 6-language coverage at launch — limited to UK/DE/IT in v1. - Final offer document generation — manual continues.

Pre-conditions (must be true before kickoff) — UPDATED 2026-06-08

• [x] Maurizio commits in writing to producing rule tables for UK Building Regs + DGNB ENV1.2 + Italian UNI norms — ✅ CONFIRMED Call 3 Q6.
• [x] Aronlight capacity gate — ✅ Aronlight v7 sent May 4; timeline compatible with EC July kickoff.
• [x] HubSpot read-only access — ✅ feasible per Call 3 Q2.
• [x] Preventivista session — ✅ reframed as optional post-spec delivery.
• [ ] Zanoni sign-off — ❓ Zanoni intro meeting June 8-9. NDA likely required before he sees full pricing. This is now the open gate.

Net status: 4 of 5 pre-conditions resolved. Zanoni sign-off is the remaining gate.

Appendix A — Materials Read (Gate)

Every artifact below was actually opened. Generic auto-summaries from the inventory script have been replaced with reads-confirmed descriptions.

Read in full / sufficient depth

Inventoried only (script auto-summary, not opened end-to-end)

Materials NOT received but expected (from call notes)

• Norwegian / Swedish sample offers — input languages but no samples. Q11 sent to Maurizio asking volume share.
• Country regulation files — Opzione A confirmed Call 3 Q6: Maurizio will write rules in our format. Legislation files as supplementary, not urgent.
• Internal supplier database — confirmed Call 3 Q9: SPI di EDM (https://edm.it/software-gestionale/). Export/API sufficient.
• HubSpot pipeline export — read-only feasible per Call 3 Q2.
• Master pod-shell library — confirmed Call 3 Q7: exists, awaiting share.
• Maurizio's Claude-rebuilt offer template — added to asks (questions-to-canta.html 2026-05-12).

Gate check — passed.

All artifacts in the "Read in full" section have non-stub summaries written from direct reads. Inventoried-only artifacts are listed with explicit reason for not opening. Critical sample (Brixton end-to-end + Karlsruhe LV preamble + Karlsruhe Angebot) is sufficient to support sections 2-5.

Appendix B — Open Questions

Resolved by Call 3

1. Master pod-shell library — ✅ Call 3 Q7 — exists, Maurizio will share.
2. K markup variability — ✅ Call 3 Q8 — 34% base + variable commerciale K (3-6%), direzione approves.
3. Supplier database — ✅ Call 3 Q9 — SPI di EDM (https://edm.it/software-gestionale/).
4. Country regulations commitment — ✅ Call 3 Q6 — Opzione A confirmed.
5. Lost-deal anatomy — ✅ Call 3 Q4+Q10 — 38% prezzo, 30% bagni in opera, 28% appalto perso. 068-26 was the capitolato example, not abandoned. 127-25 (Lucens) won, 060-25 (Karlsruhe) lost.

Still open — to ask Maurizio

Sent in the updated questions-to-canta.html (Q11 added 2026-05-12).

1. Descrittivo template canonicality — Brixton 40c × 76r. Karlsruhe 41c × 104r. 068-26 38c × 77r. Same template + project extensions or genuinely different? ⚠ PARTIAL — Call 3 Q3 suggests same canonical template. Resolve via inspection of master spec docs (pending).
2. FILE K vs Calcolo Prezzo — Are both files always produced? Will infer from master spec docs once received.
3. The "second client" Maurizio mentioned — Worth a 30-min intro call before EC proposal goes out.
4. Norwegian + Swedish coverage — ASKED in questions-to-canta.html Q11. Awaiting his answer.
5. Maurizio's Claude offer template — ASKED in questions-to-canta.html ASKS section. Awaiting his share.

Dropped from tracking (not real open questions)

• ~~Folder-structure normalization~~ — internal implementation detail. System handles either convention.
• ~~Validation gate before preventivista handoff~~ — implicitly resolved by Call 3 Q1: with 1 preventivista bottleneck + 3 back office, no formal QA gate today. Our human-in-loop UI becomes the QA gate.

Appendix C — What we did NOT do

Honest record of where this assessment fell short.

• Did not read all 5 offer folders end-to-end. Brixton (UK) read in full; Karlsruhe (DE) read final offer + 30/144 of LV; 068-26 inventoried only (later resolved Call 3 as the capitolato example); Lucens (CH) and Crissier (CH) NOT opened at all.
• Did not read Karlsruhe LV pages 31-144 — the actual bathroom-specific spec is in there.
• Did not run actual extraction prototype on the Brixton descrittivo. Schema dimensions confirmed structurally, not by attempting to extract Karlsruhe's data INTO Brixton's schema.
• Competitive scan was time-bounded (~10 min), not the 60 min the rubric calls for. No actual web searches run. Direct competitor names listed are placeholders (❓ markers). Biggest gap.
• Did not contact preventivista or back-office staff directly. Schema understanding comes from artifacts + Maurizio's narrative.
• Did not extract sample data from Karlsruhe descrittivo Excel to verify schema compatibility with Brixton template.
• Did not check Aronlight quote-gen scaffold's actual implementation depth — confirmed file existence + names, did not read the code.
• Did not run unit-economics scenarios on per-descrittivo pricing variants beyond the 4 listed.
• Did not check existing AdapttoAI work-in-progress that might collide. Lamosa SAP B1 work, BCP HR Readiness assessment, AutoalDía WhatsApp chatbot — capacity overlap with EC is asserted but not modeled. The "kickoff July 2026" timing is a reasonable guess, not a calendared resource plan.
• Did not scope the DeepL Pro license cost into the build budget — the Mes-1+ DeepL business plan (~€60-80/mo) is a recurring cost we'd pass through to client OR absorb.
• Did not red-team Maurizio's parallel DIY build risk in detail.