Eight Zones, One Enclosure

In my garage there is a metal enclosure mounted to the wall. Inside it are four power supplies, four fuse blocks, five dimmer boards, a network-connected microcontroller, a row of DIN terminals, six barrier strips, and roughly sixty feet of wire connecting all of it. This enclosure powers and independently dims the LED lighting in eight rooms of my house.

I built it in a single weekend. Nothing caught fire. Nothing shorted. Every zone worked on the first power-up, and all eight showed up in Home Assistant within minutes of flashing the firmware. This was not because I'm an experienced electrician. It was because the AI gave me something more useful than a wiring diagram: it gave me a build sequence.

The difference between a diagram and a sequence

A wiring diagram tells you where everything connects. Terminal A goes to terminal B. The positive output of Driver 1 feeds the input stud of Fuse Block 1. The data line from the controller's LED1 port connects to the D input on Dimmer Board 1. If you can read the diagram, you know what the finished system looks like.

What a diagram does not tell you is what to connect first. Or when to stop and test. Or which mistake, made in stage two, will be invisible until stage eight and then require you to unwire half the enclosure to fix.

The AI gave me a 10-stage implementation guide. Each stage built on the previous one. Each one ended with a checkpoint — a question I should be able to answer before touching the next wire. The stages were not just a convenient ordering. They were a failure-containment strategy.

The architecture

The system is straightforward in concept. Four Mean Well constant-voltage drivers convert mains power to 24 volts. Each driver feeds a Blue Sea 5045 fuse block that distributes fused positive power to two lighting zones plus the dimmer board that controls them. A QuinLED Dig-Quad microcontroller running WLED firmware sends digital control signals to four Dig2Analog dimmer boards, which convert those signals into PWM dimming on their output channels. The LED strips see constant 24 volts on the positive side and a pulse-width-modulated ground path on the negative side. The dimmer board doesn't output voltage — it completes the circuit by switching the ground path on and off thousands of times per second.

Eight rooms. Eight independently controllable lights in Home Assistant. One enclosure.

The complexity isn't in any single connection. Every wire run follows the same pattern: fused positive from the block, dimmed negative from the board, both land on DIN terminals, then out to the room through barrier strips if the zone has multiple physical runs. The complexity is in the quantity — sixty-plus connections — and the fact that a single wrong connection in the negative bus can create a ground loop that makes three zones flicker at random intervals, and you will never find it by looking at the diagram because the diagram only shows what's right.

How the sequence worked

Stage 0 was mounting. No wiring. Just screw everything to the backplane and confirm that every terminal is reachable with a screwdriver. The AI said to pause here, and I'm glad it did, because I discovered that two fuse blocks were too close together and the wire entry angles would have forced awkward bends in the next stage.

Stage 1 was the negative bus. One long screw terminal strip. Every single ground wire in the enclosure lands here: all four driver negatives, the controller ground, all four dimmer board power grounds. The AI said to pause and confirm: "You should be able to point to one location and say 'all negatives go here.'" I could. This is the foundation. If the negative bus is wrong, nothing else matters.

Stage 2 was driver positives to fuse blocks. Four wires. Nothing else. The AI said to pause. "You now have four separate positive supplies entering four separate fuse blocks." Verified. At this point the enclosure has power distribution and a common ground, and nothing is connected to anything that could be damaged by a wiring error.

Stage 3 was fuse assignment. My rule: slot 4 on every fuse block is always the dimmer board power fuse, rated at 3 amps. This rule exists because I know I'll be servicing this enclosure in two years and I won't remember which fuse is which. Consistency is a form of documentation.

Stage 4 powered the dimmer boards from their respective fuse blocks. Stage 5 powered the controller. Stage 6 connected data lines — thin signal wire, 22-gauge, carrying almost no current, physically separate from the power wiring. The AI was explicit: "Do not connect anything to the + in the +/D/- block." I would have connected something to the + in the +/D/- block.

Stage 7 was channel assignment. Each dimmer board has four output channels labeled R, G, B, and W. My strips are white-only, so I used two channels per board — W for one zone, R for the other. The AI told me to print the mapping and tape it inside the enclosure lid. I did. Eighteen months from now, this label is the difference between a 30-second service call and a 30-minute diagnostic.

Stages 8 through 10 were DIN terminals, barrier strips, and zone wiring — the actual connections to the house. One zone at a time. Test after each zone. The AI said to pause after each zone, and by this point in the build I understood why: if zone three doesn't work, the problem is in the wiring I just did, not in the infrastructure I built in stages 1 through 7. The checkpoint narrows the search space.

What the AI understood about me

The build sequence wasn't just technically competent. It was designed for a specific kind of builder — someone who knows enough to be dangerous but not enough to recover gracefully from mistakes.

An experienced electrician doesn't need checkpoints. They wire an enclosure in whatever order makes sense for the physical layout, test at the end, and fix anything that's wrong because they've fixed a hundred enclosures before and they know what a ground loop looks like when they see one. They carry the failure cases in their hands.

I don't. I carry a conversation in my phone and a checklist taped to the garage wall. The AI knew this — not because I told it, but because the way I described the project made it obvious. I wrote a PRD for a lighting system. I specified non-goals and safety rules. I used the phrase "hard to wire wrong" as a design requirement. These are not the words of someone who wires enclosures for a living. They're the words of someone who designs systems for a living and is applying that skill to an unfamiliar domain.

The build sequence is the AI matching its output to my competence level. Not dumbing it down — the technical content is precise and correct. But ordering it so that every stage is independently verifiable and every mistake is locally containable.

The rule about the fuse

One detail stayed with me. The AI's plan included a safety principle I hadn't considered: the fuse protects the wire, not the load. A 5-meter LED strip at 19.2 watts per meter draws 4 amps at 24 volts. If I'd fused that run at 10 amps because the strip could theoretically handle it, the fuse would protect the strip but not the 18-gauge wire feeding it. The wire would overheat before the fuse blew. The correct fuse size is determined by the weakest component in the path, which is almost always the wire.

This is the kind of principle that seems obvious once you hear it and that you will never derive on your own from a wiring diagram. A diagram shows connections. A principle explains why some connections are dangerous even when they're correct. The AI embedded the principle inside the sequence — stage 3 included a fuse-sizing table matched to wire gauge, not load — so that by the time I was inserting fuses, the right answer was the only answer I had in front of me.

What the enclosure does now

It powers the warm-white LED strips in eight rooms. The theatre glows at 15 percent during movie nights. Hazel's room runs at 5 percent as a nightlight controlled by a Home Assistant automation that responds to a temperature sensor. The hallway lights turn on at 20 percent when the nursery door opens after 10pm. The office lights follow a schedule. The living room dims to zero when the Lutron scene says "goodnight."

The system runs at a hard cap of 40 percent brightness, enforced in WLED firmware, because these lights are for ambiance, not illumination. The four Mean Well drivers are rated for a combined 66 amps. At 40 percent cap, the total system draw is under 20. The headroom is absurd, and that's intentional. An oversized power supply that never strains is quieter, cooler, and lasts longer than a right-sized one running at capacity.

The enclosure door has a laminated wiring diagram taped to the inside. The fuse covers are labeled. The DIN terminals are labeled. The barrier strips are labeled. The Dig2Analog boards have stickers — DA-1 through DA-5 — and the channel mapping is printed in 14-point type and taped next to them. Every name — THEATRE, LIVING, HALL BATH, HALLWAY, OFFICE, BEDROOM, PRIMARY BATH, HAZEL — appears in the exact same form on the physical hardware, in Home Assistant, and in the WLED firmware.

This consistency was the AI's idea. Use the same names everywhere. Not similar names. Not abbreviated names. The exact same string. Because at 2am, when the hall bathroom light is flickering and you're standing in the garage with a flashlight, the only thing that will save you is being able to look at a fuse, read the label, and know with certainty which zone it controls.

The system has run for months without a single failure. But if something does fail, the person who opens the enclosure — whether it's me or someone I've handed the house to — will find a system that explains itself. The labels are the documentation. The fuse positions are consistent. The negative bus is in one place. The sequence that built it is the same sequence that will diagnose it.

The AI didn't just help me wire a lighting system. It helped me wire a lighting system that a stranger could understand. That's the difference between a project and infrastructure.