From hypotheses to a working plan
The kickoff converted the v1 baseline into commitments. The engagement is closed-won, month-to-month, and active. Five of the six open decisions are settled; everything below is now the operating plan. The remaining three items are delivery-alignment choices — they refine execution, they do not gate the relationship.
Five settled, on the record
Each card is a decision from the v1 alignment set, now confirmed live. Four matched the recommendation; one diverged toward a clearer target. All are facts going forward.
The bright line holds
Carlos owns the Synology hub, on-site install, and splat training. Jonathan owns the white-label AI concept tool (client self-drives via login). Carlos walks Jonathan through the pipeline phases, handoffs, and formats. No overlap by construction.
AI pulled forward
Confirmed: site capture → AI concepts → SketchUp/TwinMotion rebuild → TwinMotion presenter EXE → Unreal finale. AI-concepts now run as an immediate parallel track — stood up as a standalone, not waiting on splatting.
White-label, client-driven
Jonathan owns the browser-based generation tool; the client self-drives via login to explore design directions. Carlos executes the chosen direction downstream in TwinMotion/Unreal. Concepts feed upstream of 3D, no overlap.
Weekly rhythm, month-to-month
All-in, month-to-month (closed-won). Weekly Thursday 3:00 PM ET recurring from 7/16; interim syncs 6/26, 7/2, 7/7. Comms run in a Google Chat "immersive design meeting" space.
Drive now → Synology NAS
Interim staging on Matt's Google Drive, then handed to Carlos's Synology NAS (in-house hub, ~20→40TB, Cloud CoWork sync). Satisfies the "in-house control" intent via the studio's Synology rather than a separate Matt server.
Closed-won & live
First monthly payment received 6/19. Terms unchanged: $10,900/mo, month-to-month, 90/10 (immersive-studio workflow / core-firm ops automation). The engagement is active.
The confirmed split
The proposed boundary became the operating boundary. Each lane now carries its real ownership scope as captured on the call.
- Synology NAS hub + on-site install
- Reality capture (Post Shot in-house)
- Gaussian-splat model training
- TwinMotion / Unreal execution + QC
- Walks Jonathan through pipeline handoffs/formats
- White-label AI concept tool (client self-drives via login)
- Inputs → ops → outputs pipeline orchestration
- Cloud CoWork automation layer
- The layer that stays aware end to end
The corrected staging flow
The one decision that diverged from the v1 recommendation, now settled with a clearer target: interim Drive, then the studio's Synology NAS as the in-house hub, with Cloud CoWork sync.
Confirmed order, AI pulled forward
Sequencing is settled with AI-concepts elevated to an immediate parallel track. The full chain runs from site capture through the Unreal finale.
New on the call, now locked
Beyond the six v1 decisions, the kickoff settled the tool stack, the capture approach, and the go-to-market track. These are operating facts.
| Area | Settled position |
|---|---|
| Tool stack | TwinMotion for live dev (presenter-mode EXE with WASD for clients) + Unreal for the cinematic finale. Lumion demoted to in-house render source only. |
| Reality capture | Post Shot in-house on the new Threadripper + RTX-6000 (96GB) workstation (arrives Mon); Lixel scanner delayed. Format VZ → OBJ/FBX mesh; splat→mesh can serve as terrain. |
| Contract | Closed-won. First monthly payment 6/19 via the aiChemist checkout. Month-to-month, $10,900/mo, 90/10. |
| Beta migration | The team is moving the two larger beta projects from SketchUp/Lumion to SketchUp/TwinMotion. |
| Marketing | The studio's marketing director produces videos (Synthesia) from project assets toward the Aug-1 sell start. |
The confirmed runway
Milestones concretized on the call, anchored by the weekly Thursday rhythm and the studio's go-to-market dates.
Three answers unblock the next gate
These are the only items still open. Each is a delivery-alignment choice with a recommended direction on the companion form. Answer all three and the next build phase clears.
Name the three folders + confirm capture overlap
The structure changed on the call to three folders: one "simple AI-beta" (small property, AI tool only) plus two larger beta projects rebuilding into TwinMotion. We need the folder names and one clarification: are the two larger betas the same projects as the 6/25 site-inventory capture, or different?
Why it matters: this unblocks the AI-concept work. Folders due ~6/26 via Google Drive + access.
Layer paid frontier models on the self-hosted baseline?
Carlos recommends paid frontier image models (plus panoramic 360) over the self-hosted SDXL baseline. The proposed posture: Carlos sets the server baseline; aiChemist layers premium stacks additively, per feature, billed as client pass-through where used.
Why it matters: it's a cost decision — we want explicit approval before wiring paid APIs into the concept tool.
When to hand GPU orchestration to automation?
Carlos wants hands-on validation of the new RTX-6000 (Blackwell) card before handing GPU/VRAM orchestration to Jonathan's Cloud CoWork automation — "if it flows nice, then we go that route." The question is timing, not direction.
Why it matters: it sets whether the first splat-training benchmark gates the automation handoff.
Three decisions, one short form
Down from six to three. The alignment form converts the open threads into a quick confirm-or-adjust pass — the recommended option is pre-highlighted on each. Pending input: Manus access + the three Drive folders, due ~6/26.
Confirm the three → Alignment form