On Running a Large Freighter - A Practical Account

What You Actually Do

You sit with it. That's most of it.

The ship handles itself - trajectory, load balance, maintenance dispatch, corridor negotiation. Your job is to know the ship well enough that when something is wrong, you feel it before the system names it. That takes about six months of continuous presence to build and never fully transfers to another vessel.

The augmentation helps but it isn't magic. The ambient layer shapes your attention - you notice things before you decide to look. The branching lets you run decisions forward and come back with something like memory of the outcome. But the layer is trained on known patterns. When the anomaly is new, it describes the problem in terms of the closest thing it recognizes, which can be worse than nothing. Part of the job is knowing when to ignore what it's telling you.

A Shift, Concretely

Most of the time you're doing something that looks like nothing.

You're reading the vessel - not checking values against range, but asking what they mean together, relative to yesterday, relative to what this ship feels like when it's fine. This is not a procedure. It's closer to listening. You develop an ear for it the way you develop an ear for anything: slowly, through long exposure, and you can't explain it to someone who hasn't.

Periodically you sit down for active work: corridor slot negotiations, cargo adjustments, tender rendezvous timing, the occasional argument with another operator about a priority conflict that happened three months ago and is still affecting your queue position. This is relationship management as much as logistics. The same people run the same routes for decades. How you've treated their emergencies is in the room with you every time you negotiate.

When you need to, you declare intent to the system: these priorities, these tolerances, this posture on the slot negotiation. The ship executes. You watch execution against what you meant, not against what you said. Those are different things, and the gap between them is where most errors live.

When Something's Wrong

The ambient layer surfaces it - a shift in your attention, a sense that something needs looking at. Your first question is whether the layer's framing of the problem is itself wrong. It usually isn't. Sometimes it is, and that's the dangerous case, because you'll be very confident in the wrong direction.

You run branches if the decision space is clear. You don't run branches if the situation is novel, because branching will return plausible-feeling futures built on the same false assumption that caused the problem. Recognizing that is the actual skill. There's no procedure for it.

Sometimes you walk to the compartment. Not to fix anything - the tickbirds handle execution. You go to see it. Smell, vibration, the specific way a surface has changed since last quarter. You come back and the system works with better information. The walk takes an hour. The improvement in model accuracy is worth it.

The failure you're there to catch is the one where every subsystem is running correctly and the combined outcome is heading off a cliff. No individual reading tells you. The trend across readings, held against your felt sense of what this ship should feel like by now - that tells you. There's no shortcut to having that felt sense. You either have it or you don't, and you only have it by being present for long enough.

When you're good at this, you'll get told you seem odd. Like you're listening to something. You are. That's the job.

You'll also find yourself overruling the system on something it's formally better at than you, and being right, and not being able to explain why in a way that satisfies anyone. Get comfortable with that early.

The ship will outlast your contract, your attention, and your patience. It has been doing this since before you arrived and will continue after you leave. Your job is to be present enough, for long enough, that the gap between its model and reality doesn't become a catastrophe on your watch.

That's it. Everything else is detail.