Design philosophy (2026-08-19): Aristide's design-led thesis — CAD interfaces are accreted button piles and users don't know what they're missing — plus Jordan's two firsthand stories (the healer whose output couldn't be modified, the industry-wide seamless-integration gap) are recorded at /design/philosophy/. It is the lens the winning layout gets executed through, and it makes the editable-everything principle binding.
Decision brief: Aristide asked for the case for each — an advocate's argument, pros and cons, and a sharpened recommendation are at the cases page.
Addendum 2026-08-19: Aristide asked what the interface looks like under the suggested 1+3 combo (Workbench base + Graft workspaces) with the research feedback included — drawn end to end at the Combo (Jobs home · Heal · Nest · Cut Plan, plus a finding-by-finding map). The three mockups below are left intact as the decision record.
Fixed in all three (the accepted direction — Jordan's picture, confirmed by Aristide): the drawing owns the screen; eight workflow tabs across the top (Import · Heal · QC · Review · Nest · Cut Plan · Post · Deliver) wearing status colours so the job state reads from across the room; contextual left toolbar; checks living ON the geometry with numbers in a panel behind them; customer's original as a toggleable gray base layer; Jobs home + folder browse + draw-from-scratch behind the ⌂ door; per-user remembered layouts. Also fixed, from the research: auto-nest as a frictionless-override draft (finding 6), repeat-one-nest-×N (finding 7), an always-visible "arcs preserved, 0 polylined" gauge (finding 5), live sliders (finding 2).
What varies is the one open decision — the panel system. Each mockup draws the same Nest-stage screen of the same job (the real numbers from build -93) under one of the three panel proposals:
| A — Workbench | B — Studio | C — Graft | |
|---|---|---|---|
| Panel home | dock slots only | floating windows | dock slots, tear-off optional |
| Second monitor (Aristide's case) | via tear-off (exception) | native | native via tear-off (remembered) |
| Adobe workspaces (Jordan's ask) | slots remembered per user | full named presets | full named presets |
| Lost-panel risk (shared shop PC) | none | real; mitigated by Reset | none — dock is ground state |
| Canvas pixels on a 60×120 nest | canvas minus drawer | full width (panels overlay) | canvas minus drawer; tear off to reclaim |
| Build cost | lowest | highest | high (most of B's plumbing) |
| Forum-evidence fit | The research rewards learnable/visual/consistent (findings 1, 3) and is silent on floating panels specifically — no verified complaint either way. Nothing in the evidence forces B; nothing forbids it. | ||
Pick A, B, or C (or mark up what to change). That closes the panel-layout question from the look record, and detailed look-and-feel design proceeds on the winner.