GameLab · Source archive

Gamelab: Relay Zero

Originally published as an Advisory Hour Substack post on 2026-10-09. This first-party site copy preserves the text, localizes public supporting media, and connects the record to its project trail.

Original post

Gamelab: Relay Zero image from the original Advisory Hour post.
Gamelab: Relay Zero image from the original Advisory Hour post.

Relay Zero has strong influences of a game I played in the 1900s.

This screenshot above is from the opening of Relay 0, a game very clearly inspired by the original Legend of Zelda on the Nintendo Entertainment System.

For this project, I wanted to get a sense of what happens when the verification process gets tighter. If I gave the model better tools and clearer guidance, could I get a better final result? Could I get more out of the total tokens spent on the project?

That seemed to be happening on a smaller scale, so I wanted to try something larger. Build something substantial, with a verification loop that could check the work as it progressed. That was the experiment behind Relay 0.

Here we have the familiar “It’s dangerous to go alone,” which has become a meme in just about every imaginable setting. This incarnation is sci-fi pixel art. The room layout and interface feel authentic. I’m less sold on the characters. They work reasonably well, but the art overall is pretty average. It was a tertiary concern for this experiment.

Still, walking in from the overworld to pick up your sword is quite nostalgic. There’s something fun about seeing that familiar opening show up in a game I brought together-it’s a little bit of the past meets the present.

Most of the levels are already there to explore in some form, which is pretty amazing. It also reveals a fun new challenge that I haven’t quite figured out how to tackle yet.

Gamelab: Relay Zero image from the original Advisory Hour post.
Gamelab: Relay Zero image from the original Advisory Hour post.

There are a few perplexing moments where the game just stops. As far as I can tell, nothing responds. The inputs lock up. It doesn’t happen in every situation, but it’s an annoying bug, and there are several others.

For an experiment, though, I think it’s about where it should be for what I was trying to make sense of.

The question now is how I, as a single person, even begin testing a project like this. AI enables such a vast scale, and this didn’t take me nearly as long to build as you might think. But now I have this whole world to explore, and I’m not even sure how big it is.

There might be a few hours of exploration here. There might be days. How much of it do I have to play through before I have a reasonable sense of whether it works?

Gamelab: Relay Zero image from the original Advisory Hour post.
Gamelab: Relay Zero image from the original Advisory Hour post.

For now, I’ll have to content myself with playing this on an airplane. I have some travel coming up, so I’ll see how it does and what weird problems I run into.

And I’ll think a little more about what it means to verify something you’ve built when the scale exceeds what you can personally explore. What does verification even look like when these worlds are so huge? I have more questions than answers in an age where answers are received so immediately, and yet are often paper thin.

Play now: https://www.ericrhea.com/gamelab/projects/relay-zero/

Connected work