手写矢量几何、浏览器原生渲染的循环插画;每件作品都要说清它是怎么动的、成本多少、什么时候该换运行时。
二十台设备一套语法:共享等距几何与材质 token、同一个 MotionValue 时钟、每台恰好一处品牌色「信号」、一行等宽读数。画廊里各台错开相位、 同一节拍;加一台 = 一个场景文件 + 注册表一行。四种画法 —— 等距、机柜与电能表的正视图、分拣线与电池模组的平面图、接触器的剖视图 —— 说明这套语法约束的是风格与驱动方式,不是画法。族是按「动作」而不是 按行业扩的,二十台各带一种:出纸、波形、逐格点亮、盘位轮扫、往复填充、 液位涨落、旋转扫掠、传送过闸、取放易主、沿路径流动、字轮进位、门的开合、 关节链、色场、光束描线、双卷耦合、分度停歇、嵌套伸缩、双稳态吸合、 阻尼整定 —— 最后一种是全族唯一一台动作出自物理响应而非目标姿态的设备。
231 labels left
layer 1 / 14
−1.7 dB
provisioning
bay 2 · syncing
battery · 85%
108° · 3 targets
#4200 scanned
tape 01 → drive
door 01 · open
pump 62% · Δ9 °C
2.9 kW
bolt 1 / 5 · 12 Nm
B4 · 46 °C
mark 82% · 1 200 mm/s
roll 91% · 180 m/min
0 out · 2.6 s takt
level 1 · 0.0 m
coil 22% · open
480 g · settling
231 labels left
二十张卡共用一个壳(标题 / 面包屑 / 状态词 / 菜单 → 舞台 + 读数 → 标题句 + 说明 → 数据区 → 页脚 + 两个动作),数据区有行表与指标块两种词汇。每张卡只写自己的状态机, 二十张各演示一种:耗材见底、进度到头、槽位阈值、限值轮换、两段式终局、 会涨也会跌的量、逐周期改写的日志、被故障抢走的主按钮、越等越久的退避、 补偿到补不动的饱和、会随输入塌掉的预测、时间自己长出来的待办、修它要停机的 漂移、本卡管不了的终局、先解锁再确认的不可逆动作、错过就得停线的窗口、 由最慢一环定死的节拍、能力缩水后够不着的那一半、同一读数两种状态的磁滞、 两头都要付钱因而没有正确答案的取舍。 所有派生文案由状态算出,不存第二份;卡内文案沿用参考稿的英文,「···」里是跳到各状态的演示入口。
231 labels left
240 labels are queued and 231 are left on the roll, so the run stops at label 232 unless a new roll goes on first
Bay 4 is rebuilding onto a spare. Until it finishes, the array has no parity left, so a second failure loses the volume
layer 7 / 14
Each pass lays one layer and the gantry rises by its height. The estimate assumes the remaining layers match the ones already down
mains · 62%
Charging while carrying the load. Runtime climbs back as it fills, so a second outage right now would be shorter than the first
provisioning
U3 is warming and will accept traffic in about two minutes. The last empty slot is reserved for drain-and-replace
#4200 scanned
One parcel crosses the curtain per cycle. A parcel that fails the read goes down the reject chute instead of continuing, roughly one in 7
−2.2 dB
The trace sits well inside the window until the compressor kicks in, then runs three decibels over for about forty milliseconds — long enough to hear, short enough for the meter to miss
000° · 3 targets
Each revolution re-measures all three returns. Nothing is inside the 6.0 m keep-out radius
tape 03 → drive
The picker mounts one tape per pass. A tape that fails is retried with a growing delay before it is skipped
pump 50% · Δ9.0 °C
Load is climbing but Δ has not moved, because the pump is speeding up to match. A flat temperature here means spare capacity, not a quiet machine
2.9 kW
The forecast divides what is left before the next tier by the current draw, so it is a projection and not a promise. Raising the load collapses it immediately, which is the point of watching it rather than the meter reading
4 of 10 held
Parcels age one day per pass and are returned after 5 days. The oldest one here has 1 days left, which is the number worth watching
5 / 5 · 0.19 mm
The tool centre point creeps as the cell warms up, about 0.014 mm per part. It is fine until it is not, which is why the number matters more than the parts count
peak 44 °C
Temperature is spread across the pack rather than piled into one cell, which is the shape you want. The number to watch is the peak, not the average
spiral-24 · 1284
A mark is permanent, so changing the format splits the batch into a before and an after. That is why the button arms first instead of firing on the first press
roll 42% · 180 m/min
The supply roll shrinks as the take-up grows, so the same web is always between them. The splice window opens on its own at 22%
takt 4.7 s
The table cannot index until press is done, so nothing else on the list changes the takt. press is 1.2 s slower than drill — cut more than that and drill becomes the constraint instead
level 1 · 0.0 m
All three stages extend together, each covering a third of the 5.6 m lift. Anything on the queue can be reached, so the only thing setting the rate is the round trip
coil 100% · closed
It stays closed all the way down to 38%, far below the 72% it took to pull in. That gap is deliberate: without it a supply wobbling around the pick-up would chatter the contacts apart in a few hundred operations
503 g
The filler's spread is fixed; the setpoint only slides the whole distribution. Two grams up leaves 1 short at 5.3 g a pack, two grams down leaves 4 short at 2.3 g. Nothing on this dial gets both numbers to zero