We changed the front door.

Not the house. The front door.

For the first few weeks, Ship introduced itself through the thing we could prove fastest: a CLI, a catalog, a handful of versioned artifacts, and a repo you could inspect. That was honest. It was also the wrong first sentence.

The people who buy the risk are not asking "which command installs the rules." They are asking something simpler and harder:

Can I see what is moving?

Can I see what is blocked?

Can I tell who decided?

Can I show the evidence to someone who was not in the room?

That is not a CLI problem. It is an ownership problem.

The old front door was true and misleading

The old story was not fake. Ship still has technical contracts underneath it. Repos matter. Tracker bindings matter. Versioned procedures matter. Agent rules matter. The API that serves the same text to machines and humans still matters.

But the order was wrong.

We were asking a product owner to walk through the garage because the garage had the cleanest tools. "Start with the CLI" made sense to the builder. It did not make sense to the person trying to decide whether AI-assisted delivery was safe enough to put near a real backlog.

The first visible object should not be a command. It should be the workspace.

A workspace says: here is the product area, here are the repos, here is the tracker, here are the policies, here is the knowledge agents are allowed to use, here is the Inbox where decisions land, and here is the evidence trail.

That is the product.

The CLI, where it remains, is plumbing.

A correct internal model is not the same thing as a correct product story.

We had the internal model. We were leading with it.

Why product owners first does not mean engineers second

There is a lazy version of this pivot where you hide the engineering truth behind a friendly dashboard and hope nobody asks what is underneath. We are not doing that.

Product owners first means the public language starts with ownership. Engineers still get the contracts. They just no longer have to watch us pretend those contracts are the buyer's first problem.

The owner sees:

which work is blocked;
which policy applies;
which evidence exists;
which human needs to decide.

The engineer sees:

which repo is bound;
which workflow ran;
which PR carries the change;
which rule or procedure shaped the work.

Those are not two products. They are two views of the same record.

The mistake was treating the engineering view as the entry point because it was more concrete. Concreteness is not the same as relevance. A product owner does not need less truth; they need the truth at the altitude where they make decisions.

The new first mile

The new first mile is deliberately boring.

Create a workspace. Connect GitHub. Bind the tracker. Set policies. Seed knowledge. Review the dashboard and Inbox.

There is no magic in that list. That is the point.

The magic-sounding version of AI delivery is: "agents will write code for you." The useful version is: "repeatable work can move faster if the boundaries and evidence are visible."

So we moved policies earlier.

Policies are not legalese. A policy says what an agent may do, what evidence it must leave, and when it must stop for a human. That is the missing middle between a prompt and a process.

Without policies, every agent run becomes a little exception. With policies, the run has a shape before it starts.

What we deleted from the story

We deleted tools and collections from the public landing path.

Again, not because the repository never had those nouns. Because they were internal packaging nouns leaking into the product story. A buyer does not care whether an integration descriptor used to live under artifacts/tools. A founder does not care whether a starter bundle used to be called a collection.

They care whether the workspace can answer:

what can the system observe;
what can the system change;
what needs approval;
what proves the change worked.

If a noun does not help answer one of those, it does not belong on the front page.

We also removed setup-through-terminal as a starting path. There may still be technical reference for teams that need local control. But no one should meet Ship as a command they must copy before they know what they are buying.

That is how open-source infrastructure talks. It is not how an owner buys safety.

The awkward part

Any visible positioning change risks looking like a reversal. "Yesterday you said CLI. Today you say product owners. Which one is real?"

The honest answer is: both were real, but only one was the right first sentence.

Early products often over-explain the part their makers understand best. For us, that was the implementation contract: files, commands, manifests, adapters, protocol. We did not invent that story to sound technical. We started there because it was the thing we could audit.

Then the console got real enough that the owner story became more true than the implementation story.

That is not a brand pivot. That is the product catching up to the reason it exists.

Ship exists because AI-assisted delivery without human ownership turns into a pile of unexplained activity. The antidote is not a better command. It is a workspace where intent, policy, work, and evidence stay together.

The technical substrate still earns its place. It just does not get to be the door.

The principle

Lead with the person who owns the consequence.

The engineer owns the implementation. The product owner owns the decision. The founder owns the risk. If the product starts at the implementation, everyone else has to translate before they can trust it.

We moved the front door so the first thing you see is the consequence: work moving under visible policies, decisions landing in an Inbox, evidence attached to the record.

The garage is still there.

You just do not enter the house through it anymore.