25 · Project Variations | Generative AI for Game Development

The last video in the series asks whether an engine built for games can do anything else. Its answer is that it does whatever the goal config says: educate, persuade, comfort or entertain. This chapter introduces no new projects. Instead it offers new purposes for the system the reader already built, a modular engine made of JSON scenes, GPT expansions, decision logic, emotional scoring and procedural chains. It accompanies the book by Shesh Narayan Gupta and is narrated by Liam for Humanitarians AI.

Six terms for customizing a workflow

The video defines six terms. A workflow is the pipeline from input to story. Domain vocabulary is the set of words a field needs. The outcome is what the story is meant to do: learn, persuade, comfort or entertain. Personalization means fitting stories to the person. Collaborative intelligence treats AI as a partner rather than a replacement. Ethical design means building in reflection at every step.

Education

The Flask and JSON engine can become a language tutor, with vocabulary hidden inside plot choices. It can also become a history simulation in which the learner's decisions fork the timeline.

Marketing

Examples include a skin care quiz that makes the user the hero of a self-care journey, emails that evolve with the recipient's replies, and loyalty paths that reward engagement. The video describes this as broadcast becoming dialogue, and it warns that the personalization line must hold.

Wellness

A therapeutic journaling assistant uses the same engine with slower pacing, affirming logic and gentle words tuned to comfort.

Customize the inputs, not the engine

Customization means changing the inputs, the vocabulary and the outcome while the engine stays. A teen slang chatbot and a corporate simulator share nothing but architecture. The video shows a sketch, honestly labeled as such, in which one traveler at a crossroads is educated, persuaded and comforted by changing only the goal config. The recommended method is to define the goal first, whether that is to calm, excite, educate, provoke or heal. Then iterate by testing against user reactions, observed behavior and ethical review, because no config is final.

Three directions forward

The video closes with three directions: greater personalization, collaborative intelligence and ethical design. Its framing is that generative storytelling is not just technology but a craft shaped by human intent, and that the pen now belongs to the reader.

Key takeaways

• One modular engine can serve many domains.
• Education, marketing and wellness are three adaptations of the same Flask and JSON engine.
• Change the goal config, vocabulary and pacing, not the engine.
• Define the goal first, then iterate against real user reactions.
• Marketing personalization turns broadcast into dialogue but must be handled carefully.
• No configuration is final; test it against behavior and ethical review.
• An ethical line should be specific to the domain, not a platitude.

Try it yourself

Paste the video's prompt into Claude and name one non-game domain. Ask for a goal config covering the outcome, ten domain vocabulary words, the input format and the pacing, then one scene running under it, one ethical line you must not cross and a way to test the adaptation. Check that the ten words are ones a practitioner would recognize, and that your ethical line is specific to this domain. If it could apply to anything, it protects nothing.