Why we publish our prompts

Transparency, at most studios, arrives as a single line buried in the credits. We think that’s backwards. If a tool shaped the story you just played, you deserve to know where — and so does the next person trying to build something similar.

So every Ometepe project ships with a research log: a plain-language record of which parts of the work were AI-assisted, AI-generated, or AI-enhanced, what we tried that failed, and the seams we deliberately chose to leave visible. The log isn’t a disclaimer. It’s part of the work.

What the log records

Three columns, roughly. The first is provenance — what was generated, with which tools, under what direction. The second is authorship — the decisions a human made before, during, and after. The third is the part most write-ups skip: the failures, the dead ends, the moments the system produced something convincing but wrong.

A method you can’t inspect isn’t a method. It’s a trick.

There’s a selfish reason for all this, too. The logs are how we get better. Reading back through a year of them, patterns surface — which prompts reliably lost the thread, where a human pass was non-negotiable, which forgotten genre turned out to fit a generative pipeline like a glove.

We publish the prompts, too. Not because they’re precious, but because the opposite — hoarding them — assumes the technique is the value. It isn’t. The thesis is the value. The prompt is just how we tested it this time.

If you’re building in this space and want our log template, it’s open. Write to us.