The case for each interface — pros and cons

← the mockups · A · B · C · the Combo · the research

Asked by Aristide 2026-08-19. What is being decided is ONLY the panel system — Jordan's canvas-first basis, the eight status tabs, checks-on-geometry, the Jobs door and everything the research fixed are identical under all three. One clarification first: the Combo page is not a fourth system. It is C realized — the Graft drawn end-to-end with A as its everyday ground state. So there are three real candidates, and each gets an honest advocate below.

The research findings cited in this brief (from the verified operator research):

A — the Workbench (docked slots)

"cadmaster is a production tool, not a creative suite. The laser doesn't rearrange its own control panel, and neither should the software in front of it. Every operator on every shift walks up to the same screen; every support call starts from a known layout; every hour NOT spent building panel plumbing goes into what the research says actually wins — geometry fidelity, manual nest control, learnability, and a visible improvement cadence. LightBurn took an entire market with no floating panels at all. Ship the workbench, and let real users EARN the fancier machinery by asking for it."

Pros

  • Cheapest and fastest to a solid v1 — and finding f3 (usability beats feature count) says users forgive what's missing when the core is good, so speed here buys goodwill everywhere else.
  • Least plumbing = least bug surface. Finding f4 says reliability failures — not aesthetics — are what make operators write software off; every moving part we don't build can't break.
  • Identical on every workstation: training, documentation, screenshots, over-the-phone support ("look at the top right") all trivial. "Where did my panel go" is structurally impossible.
  • Novice-proof for a shared shop-floor PC and rotating operators.
  • Canvas gets a deterministic, predictable share of the screen — the 60×120 sheet never fights a float for pixels.

Cons

  • Aristide's stated want — "I decide how my real estate works… toolboxes on the extra screen" — is served only through the tear-off exception, and the slot system is a hard ceiling on it.
  • Jordan's Adobe-preset instinct shrinks to "which panel sits in which slot."
  • If usage later demands workspaces anyway, retrofitting them into a slots-only codebase likely costs more than building C now.
  • Weakest demo/sales moment of the three — it looks like every other utilitarian tool.
Pick A if: v1 speed and shop-floor predictability outrank personalization — many hands, shared machines, tight timeline.

B — the Studio (free-floating Adobe workspaces)

"The people who will sit in front of cadmaster all day are CAD people — the same people who live in Illustrator and AutoCAD, where panels are windows and the workspace is yours. finding f8 says what professionals love is automation of the tedious PLUS fine manual control; screen real estate deserves the same respect as geometry. Multi-monitor is the norm at every programming desk, and B is the only option where the second screen is native rather than an exception. Build the studio and cadmaster never hits a ceiling — no future panel, no future monitor, no future workflow the layout can't absorb. This is what 'a real pro tool' looks like, and it's the version people show their friends."

Pros

  • Total real-estate control — Aristide's extra-screen case is just where windows live.
  • Workspaces at full power: named per-user, per-stage layouts; switching tabs can switch the whole arrangement. Jordan's Adobe muscle memory honored completely.
  • Highest ceiling: tabbed panel clusters absorb every future panel without redesign.
  • Strongest demo and sales impression; feels premium.

Cons

  • Heaviest build by far — window management, snapping, clustering, persistence, multi-monitor edge cases. Finding f4 (reliability failures are what kill) cuts straight against it: the plumbing IS the bug surface, and flaky UI is the cursed failure mode.
  • The classic shared-workstation failure: a panel "disappears" for the next login, or floats over the geometry someone is reading. Mitigations (per-user workspaces, Reset) are yet more plumbing.
  • On a single small screen it is strictly worse than a dock.
  • The research verified NO operator demand for floating panels. The whole case rests on the partners' own taste — real, but it is taste, not evidence.
Pick B if: the daily users are few, expert, and own their workstations — and personalization is part of what you're selling.

C — the Graft (dock ground state + tear-off + named workspaces) — the Combo as drawn

"The real question isn't dock versus float — it's what the GROUND STATE is. C answers: the ground state is A. Walk up to any workstation and it's the workbench — consistent, learnable, unlosable, everything the evidence actually rewards (f1 learnability/visual, f3 usability over features). Freedom is then an explicit act: tear off the panel you want, put it on the monitor you want, and the workspace remembers it for YOU without changing the floor's default for anyone else. It is the only candidate that satisfies both partners' stated wants at once — Jordan's Adobe presets AND Aristide's real-estate control — while making the one famous failure of floating UIs structurally impossible, because the dock is always home and snap-back is always one click. And it ages honestly: a shop that never tears off is simply running A; a programmer who lives floating is running most of B. The interface meets each user where they are, which is the product's whole philosophy."

Pros

  • Both partners' verbatim asks served natively; second monitor first-class and remembered.
  • Lost panels impossible — every float has a home slot; Reset restores the workbench.
  • Evidence-aligned default (consistency and learnability, findings f1/f3), taste-driven freedom strictly opt-in.
  • Covers the widest range of future customers — shared floor PCs AND single-seat programmer desks — without two products.

Cons

  • Costs most of B's plumbing (tear-off windows, workspace store, snap logic) — close to B's price without B's full freedom.
  • Two modes to build, test and document; the subtle bugs live exactly at the docked/floating boundary.
  • Floats can't form free mid-air clusters — the dock template defines what exists; a hard-core B user will still bump the ceiling.
  • Slowest of the three to a COMPLETE v1 if all machinery ships day one (mitigable — see verdict).
Pick C if: you want A's floor discipline and B's personalization together and accept paying most of B's build cost for it.

Verdict — my recommendation, unchanged but sharpened

C, built in A's order. Architect the panel framework for tear-off and workspaces from day one (that decision is nearly free at design time and expensive to retrofit), but SHIP the docked workbench first and turn on tear-off + named workspaces as the fast follow. That way v1 arrives at A's speed, the research-backed default is what every new user meets, and the partners' personalization lands weeks later instead of being either abandoned (A) or paid for up front in schedule (full C) — while the visible improvement cadence operators reward (finding f1) gets an early, meaty release.

What would flip me: to plain A — if the realistic customer is mostly shared workstations with rotating operators and v1 date is sovereign. To full B — if the realistic customer is single-seat programmer desks like Aristide's own, where layout freedom is a selling point and nobody shares a login. Worth deciding on that fact, not on taste: who actually sits in front of cadmaster at customer number ten?

Whichever letter wins, the choice gets recorded as a ruling on pin 7b and detailed look-and-feel design proceeds on it. Nothing here is built — phase rule stands.

← the mockups