SCALABLE, AI-ASSISTED DIGITAL GAME PLATFORM

I led Claude Cowork and one lead developer to create a generative trivia database with automated curation layer. The result is a self-growing party game engine with application across mobile, video, and immersive experiences – truly novel and valuable IP.

The goal: Can a team of two provide a daily mobile trivia game with wordplay and whimsy?

Trivia is intrinsically a curated game with consumable, one-time-play puzzles. This is in contrast to platformer and casual match games. Automated curation of generated content is no small feat – but is it now possible with Anthrop\c? Let’s find out.

In order to design interfaces with AI, you need to know what you want and why.

Leading up to prompting a machine to “make the thing for us” were many cycles of ideation, preparation, and iteration to define the product: proof that AI is not the creative resource – it enables the creative resources to move forward more quickly.

AI doesn’t do everything well. Delineating roles, defining stories, aligning goals, and creative control – these are what the human stakeholders bring to the table. Such was the case here: What is the game being played? Is it fun? Where is it not fun? The emotional and behavioral aspects of even the simplest game are not the wheelhouse of an LLM. Human beings still need to guide these aspects.

A ket aspect of this game is humor – another weakness of AI. Its generated puzzles lacked the proper timing and nuances of wordplay humor. This game blends trivia and dad jokes – a rather simple humor device that the AI still struggles with. But there are reasons things work or don’t and all we needed to do was find those reasons, and communicate them as rules to the bot. Only then can we trust the bot to take over the job.

Knowing I wanted to end up at largely automated puzzle curation for scalability, I assured there were human-attended gates in early stages, so that we could see what will lead to quality automated curation of these humorous puzzles, tweaking the generative layer for maximum humor, and creating curation bots to recognize the rules of the road.

Once those rules were established, one puzzle type after another, that automated curation can take place.

Only then can we trust the bot to take over the job.

This means a creative resource, whether human or ultimately AI, needs to be able to effectively communicate the mechanical rules behind the function of the product – and that will always be true.

Developers Are Better At Creative Works

Telling an AI to tweak something can take more time than you’d assume. Giving a developer creative license is a better bet, as they can more easily navigate choices in converting rudimentary, functional AI into something with style and novelty.

AI has changed how we created interactive products. But we will always need a developer and a creative - to provide guidance and assure quality. Humans are still the best choice for filling these roles.

A human curation layer helps us see and define what automated curation would need to be effective.

Testing the database with a rudimentary, but thorough, AI-generated mobile interface

Reviewing the generative design and packaging it up for handoff.

Trusting the developer more than the bot to translate the rudimentary interface into something with novelty and style.

Next
Next

InMarket • Automation Victory