Terrava Earth

Inspiration

Fascination with history, and how much of it happened at the same time. In particular the history of China, which I was less familiar with. It is amazing rich (I saw the Terracotta warriors last summer; fascinating on many levels).

But it's hard to fathom simultaneous history. So having built a map app with AI coding previously, and thinking what would be ambitious for this hackathon, and ready with the release for 5.6 Sol et, this seemed like a prime candidate to try.

I also had the idea that this could be a good local build & host split. As in all the hard work could happen locally with Codex as the engine harness so to speak... I with I had more time, because I think this is going to work really well over time.

What it does

A global interface for exploring power, place, and time across every polity in recorded history.
A large globe-centric view.
Power, place, and time journeys. Selected polity details.
Timeline across history with changing polities and views.
A local agent engine continuously expands, refines, and publishes the atlas.

How I built it

Codex, obviously. 99%. The rest ChatGPT.
I have my own detailed plans that I start with. I still mostly like them better than built-in /plan (but that is changing).
I started with big /goals (3 days). Then significantly refined from there.
I heavily use robust, strict linting, preflight checks. Has proven VERY useful for AI coding.
I heavily use my own about a dozen review skills (bugs, DRY/SSOT, evergreen docs, 'make it right'/'make it fast'. mapbox, etc.) Again, VERY helpful for AI coding.
All of that compounds over time. But admittedly starting out overly ambitious make this more of a grind and not enough product tuning in the end for the hackathon.

Challenges we ran into

Ironically, Sol Ultra. Being so new and encouraged by the opening discussion by Greg et al, I dove in full bore. Two projects, planned with Ultra, deep plans, turned into building through /goals, again both with Ultra.
Many many tokens later, and about 3 days, I more or less stopped both goals (prior side conversations were not quite enough to tame them). I had two semi prototypes. Terrava Earth, despite the weaker prototype, I felt had the better chance.
There is no way I'll get all the polities I wanted to in time. Even with the basic architecture of local build to save time, this will take probably another week or more to get all polities in place.
The second challenge is I really wanted this live. I think hard to judge if I had to instruct how to setup. The Cloudflare process with the local build assets and hosted version were more challenging than I expected. I am still not 100% sure this will work all the way by deadline. We will see on that as well!

Accomplishments that we're proud of

I proved the idea. This will work.
The agent engine, while feeling a bit sloppy, I think will really be powerful once underway for real and cleaned up a bit.
Another map app; these are still hard.
The app is fun! And will get more fun. Educating and fun. Great combo.

What I learned

The agent engine as mentioned should be a very useful tool, both in general, and for the local/host split.
And specifically to the prior point, Codex as part of the stack, not just coding environment.