Won the job? The nesting is already done.

The rule layout. The same rule under a second seed. A two-way rotation. A modestly tighter gap that still sits at or above the family floor. And a conjoined pass that puts parts on a common line. Every racked offcut gets a candidate of its own on top of that, plus one sweep across the whole rack. On the plate above that came to seven attempts, and the page says so rather than making you take the first answer on trust.

Then it chooses, and the order it chooses in is the part worth knowing: fewer sheets first, utilisation second. A layout that is a percentage point tighter across three sheets loses to a layout that fits on two, because the second one is the cheaper job and the first one is the better-looking number.

The gap between parts, the margin at the sheet edge and the width of a micro-joint come from the material family and the thickness. They are not three numbers you set once and then carry onto every job after it. Sixteen gauge mild steel and half-inch aluminium do not want the same spacing, and a nester that asks you for one number is asking you to remember which job it was set for.

On the plate above the rule gave a 3.2 mm gap between parts and kept 5 mm clear at every edge. It also declined to conjoin anything: aluminium's ceiling for common-line cutting is 3.0 mm and that plate is 3.175 mm, so the control is greyed out and the footer reads CL SAVED 0.0 mm. Forcing it on would override the metal's own rule, which is the opposite of what this section is about.

Every value is overridable, because your shop knows things a rule does not, and changing one pins it for that job rather than for every job after it. The margin has a floor of one millimetre, which is the one number the rules will not let you argue with.

Writing a program draws the stock off the rack once and records it against the job. Re-printing the same program does not draw it twice, which is the failure mode that makes inventory counts drift until nobody trusts them. Whatever usable material is left goes back on the rack with its own tag.

The offcut book has to balance: parts plus racked offcut plus scrap equals the area that was opened. When it does not balance, the report says so rather than rounding the difference away.

Cycle time is not an estimate typed into a field either. In the product's own words on that page: cycle time is the compiled program's own estimate, so this page cannot disagree with Sequence or Program about it.

That amber line is the whole policy in one sentence, and it is the page's own: a rate that is not set shows as a gap rather than as free time. A zero in a total is how a job goes out at a loss and nobody finds out until it ships, so there are no zeros standing in for figures nobody supplied.

A solved nest carries its own provenance: the seed it was solved under, how many generations ran, how many placements were evaluated, and two digests, one of the inputs and one of the result.

That is not bookkeeping. Hand the same nest to another machine and it re-materialises in a single evaluation, or it refuses outright. There is no third outcome where it quietly produces something close and nobody notices the parts moved.

The kernel is single-threaded and deterministic given a generation ceiling, which is what makes that promise keepable. The search around it is bounded by the clock instead, and you can see the bound on the plate above: forty seconds per material group, seed 1. A slower PC gets through fewer candidates in forty seconds. Whichever one it chose is recorded, and that one replays exactly.

The tube nester states the same thing on its own panel: the same seed and the same parts always give the same nest, so a result can be reproduced when somebody asks why a part landed where it did. That question gets asked about a week after the job ships, which is exactly when nobody can remember.

Before any of it is written, the Program page shows exactly what each layer will be set to. The job below never got written: the gate refused it, for reasons the next section goes through. The layer map is still there to read, which is the point of showing it here.

On your particular machine. Two posts exist, and neither has been tested against your controller. That is deliberate rather than an omission I am hoping you will not notice: I do not have your machine, and a compatibility claim I cannot stand behind costs you a scrapped sheet. What happens instead is that every program is read back after it is written and verified against the geometry it came from, which is a check I can actually run. Email me your controller and I will tell you straight where it stands.

Micro-joints are the one thing G-code cannot carry. There is no way to express a deliberate gap in a continuous contour, so the post drops them, counts them, and declares it. A micro-joint that was the only thing holding a part down is called out by name rather than left for the operator to discover.

Two things are holding it. The flat patterns contain features down to a tenth of a millimetre, and the finding says what that means rather than flagging a number: the shortest feature is 0.11 mm against a 0.16 mm kerf, so the beam will remove the whole feature. And the technology library has no profile for this metal, which is the one worth reading in full:

Nothing in the technology library matches 5052-H32 3.175 mm, so this part would be cut with "AL 6061 0.125" · air" speed, power and gas. That is a guess, and a wrong guess on the wrong alloy scraps the sheet.

It then says the part nobody says out loud: the cycle estimate and every cost figure are computed from the substituted profile too, so they are wrong until this is fixed. That is why the $429.00 in the report above is a number with a condition attached, and the product is the thing telling you so.

The page states its own rule at the top: releasing is refused while anything blocks, and the chip is green only when the list is empty. Green means clean, not merely permitted. A warning you can click past is a warning everybody clicks past by the second week.