Building a Gemini VIII narrative game: making uncertainty playable

I wanted to explore a particular kind of responsibility: making a decision in Mission Control when the people at risk are somewhere you cannot reach continuously.

That idea became Failure is Not an Option, a narrative flight-director game. The first Gemini VIII demo is now playable at finaogame.com.

I'm Dan Lee-Odinson, a Technical Project Manager and Space Studies student at Everglades University. This is a personal project and a tribute to the people of the space program. It's also a way to put my interests in Mission Operations and Program Management into a concrete piece of work that others can inspect and challenge.

One playable scenario before a much bigger campaign

The long-term vision runs from Gemini toward Artemis, with decisions carrying across missions. The current release is much smaller: one Gemini VIII scenario.

You prepare, ask controllers questions, consider the return options, and work through a debrief. The consequences reach a Gemini IX-A preparation-plan screen. That is an endpoint showing continuity, not a second playable flight.

Holding that boundary is part of the project. A campaign pitch can be expansive; a release needs to say exactly what someone can play today.

The hardest design constraint was historical responsibility

A real mission is not a branching game. The historical crew made their own decisions, and gaps in communication matter.

The demo preserves a crucial boundary: Glen Kurtz, the fictional director, does not remotely take over the crew's actions during the loss of contact. The game presents reports, questions and decisions within its authored scenario. It also identifies where it departs from the record.

For the named Houston CAPCOM, Jim Lovell, the content approach is constrained: procedural calls and sourced relays, with provenance, rather than invented personal opinions. Fictional controllers provide room for authored disagreements.

That creates an interesting problem for the interface. A historical choice label cannot imply that all of the consequences which follow are historical. The explanation has to distinguish a real timing anchor from a fictional risk model.

The NASA Gemini VIII overview provides the mission context. The game's History panel goes further into the sources and departures used for this scenario.

Consequences need to remain understandable

Preparation changes the conditions for later choices. The return decision has two authored routes, and different preparation choices lead to different result tiers. Relationships and lessons from the debrief then constrain the follow-on preparation plan.

This is a simplified narrative model. It is not an operational training simulator, and its fictional tradeoffs should not be read as findings about what the real crew or ground team should have done.

The useful question in testing is whether a player can explain a result afterward. If they cannot connect a consequence to the information and choices they saw, the design needs work.

What review found that desktop playthroughs missed

A live-site review completed both return routes and all six authored result tiers without uncaught JavaScript errors. It also exercised manual browser saves, JSON export/import and persistence of the next preparation plan.

Then the viewport narrowed.

On a phone-width screen, a large active-speaker portrait left too little space for a dialogue line that contained another portrait. Text was clipped. Separately, the toolbar refused to wrap and crushed the status information beside it.

Those are presentation bugs, but they undermine the central mechanic: reading enough information to make a decision. A working state transition is not sufficient if the player cannot see the evidence.

The findings were handed back with screenshots, reproducible dimensions and specific acceptance checks. That is the kind of project-management work I want this portfolio to show alongside the finished screen.

How AI fits into the work

Generative AI is part of the development process, including coding, content and visuals. I set the direction, review outputs and decide what is accepted. The workflow uses separate content and implementation handoffs, source notes and browser checks.

Calling it human-in-the-loop does not make the output automatically accurate. Historical claims still need sources. An attractive image still needs inspection. A generated implementation still needs to be played and tested.

This article was drafted with generative-AI assistance. The game's disclosure is: “This project was created using generative AI within a human-in-the-loop process, with human creative direction, review, and final decision-making.”

Try the demo

Play Failure is Not an Option. Desktop is recommended for a first playthrough.

Which decision made you pause, and did you have enough information to understand the tradeoff? I'm especially interested in observations from people working in operations, human factors, history and narrative design.

The full game is coming soon, with no release date announced. The wider campaign systems remain a long-term ambition.

This is an independent, not-for-profit homage. It is not affiliated with, authorized, sponsored or endorsed by NASA. It includes fictional dialogue, simplified systems and alternate historical outcomes.