01 / Overview
One operating picture for a fragmented shop floor.
CNCLux is a CNC production workspace built around the realities of scenic fabrication: drawings change, programs move, two machines produce overlapping histories, and the most important evidence may still arrive as a marked-up sheet of paper.
Its HYDRA interface brings job files, EnRoute sources, DNC exports, machine history, cut-sheet intake, remaining-work forecasts, production review, and daily notes into one navigable system. The workspace reads the existing CAD, DNC, and machine-history sources without renaming or rewriting them.
02 / Problem
The data existed. A trustworthy production story did not.
A machine log can prove that a program ran, but not that the correct revision produced usable parts. A manifest can describe requirements, but not whether sheets were consumed as planned. A folder can look complete while hiding renamed files, missing runtimes, material conflicts, or provisional halted runs.
The design challenge was to reconcile those partial truths without flattening uncertainty into a falsely confident number. In HYDRA, unknown stays unknown, a run is not automatically a good cut, and every correction remains traceable to its original evidence.
03 / Process
Design the boundaries first.
The architecture separates requirements, production evidence, material consumption, planning overrides, and operator notes. Program identity is exact—show, effective element, and complete filename stem—so similarly named files cannot quietly cross job boundaries.
Live, saved, offline, and unknown states are visible in the interface. Cached snapshots make the workspace useful when network drives are unavailable, while protected local state stores mappings, reviews, and planning choices with validation and recovery safeguards.
Read source systems; do not rewrite them.
Preserve uncertainty instead of converting it to zero.
Keep automatic evidence and human judgment distinct.
Make every correction reversible and attributable.
04 / Solution
From scan to forecast, the operator stays in context.
Cut-sheet intake pairs the source PDF with a compact always-on-top review window. Confirmations become durable evidence, material tickets can be exported, and unresolved pages remain visibly unresolved.
The same evidence flows into remaining-work forecasts. HYDRA combines planned cycles, reviewed production, current CNC exports, material requirements, and explicit capacity assumptions, then shows what is known, what is provisional, and what still blocks a finish estimate.
05 / Outcome
A practical system for planning with evidence—not optimism.
CNCLux turns daily production into an auditable loop: inspect requirements, watch the machines, review physical evidence, reconcile exceptions, forecast what remains, and hand off the next decision with its context intact.
The current engineering baseline was exercised by a 146-test regression run on September 26, 2026. Its CNC runtime model was calibrated against 298 retained runs across 212 programs; on that dated sample, the deployed model reported 0.52-minute mean absolute error and 10.0% mean absolute percentage error. Those figures are validation evidence, not promises about every future job.
Dated validation / September 2026: 146 regression tests in the full selector; 298 retained machine runs in calibration; 212 distinct programs represented; 0.52 min full-sample mean absolute error.