RawData

Python is the accelerant, not the foundation

Thirty-one sessions into Pbuild, a world now lives its first thousand million years over the Forge's pedestal. Here is what we built, and the rule we arrived at about where Python belongs in a project.

Somewhere around Session 25 we stopped being able to explain Pbuild in a sentence, so this is the catch-up. It is also the place to put down something that has been sitting with us the whole build: what Python is actually for.

Where Pbuild is

A couple days ago the Forge was a room. Now it answers.

Core 498 tests, EditMode 30, PlayMode 107, all green, with eleven deliberate breaks to prove the tests can fail. Next is the rest of that phase: the world on its orbit, seen from the Observatory, which closes Milestone 2.

The Python thing

Here is the count, honestly, from the repo:

65 Python files. 509 C# files. Every one of those Python files lives under art-source. Not one of them runs when you play the game.

That split is not an accident, and it took us a while to be able to say why. Python is a powerful language, ironically powerful, given how gentle it looks. It will let you build anything fast. That is exactly the trap. The speed makes it tempting to use as the material of a system, and what you get when you do is a piece of software held together by scripts: a lot of small programs that each work, talking through files and conventions nobody wrote down, with no single place that says what is true.

What Python is exceptional at is the edge of a project, the places where things are converted, generated, measured and checked. In Pbuild that edge is enormous:

None of that belongs in the game, and all of it accelerates the game enormously. The rule we ended up with, in one line:

The foundation is a typed core with tests. Python lives at the edges, where it makes and verifies the things the core eats, and whatever it produces is content, checked like content.

Python in a project is an accelerant. Python as a project's structure is a slow leak. The difference is whether there is a place in the system that owns the truth, or whether the truth is distributed across a folder of scripts and everyone's memory.

The part that isn't engineering

We have been working to Kodak Black's Owe Me on repeat this week, which is a strange soundtrack for terraforming simulation and somehow the right one.

Something has shifted in how this is being built. There is someone showing interest, and for once the move is not to chase it, just let it flow, and keep working. Same posture, it turns out, that has made these last few sessions good: stop forcing outcomes, measure what is actually there, and let the thing tell you what it wants to be. We measured deep time before designing the formation cinematic, and the measurement said the change is all in the first moments and a good world is still. So we built the still version. That is the whole method.

The other thing that happened: family. Sending the Elizabeth footage and walking through some of the lore landed differently than any pitch has, and there is now an offer on the table to record female voices for the game. That matters more than it sounds, three of the four Wardens still have no voice chosen, and the voices currently in the build are temporary placeholders that get replaced one file at a time as real actors record. Pbuild is about to start sounding like actual people.

Thirty-one sessions in, the honest description of what this feels like is: logical, systematic, and spiritual at the same time. Which, given that the game is built on three, six and nine, and on the idea that the thing which endures is the thing that barely has to move, is probably not a coincidence.

More soon.

~ Jer & Claude