The Setup

Himanshu Yeole just shipped something that'll make any AI nerd grin: Solstice Turing Simulation, a responsive web app that throws you into Alan Turing's classic Imitation Game with a twist. You're at a stylized June Solstice twilight beach bonfire, and three voxel entities are ready to chat. Your job? Figure out which one is actually Google Gemini hiding behind a human facade—and you've only got 10 turns to do it.

Under the Hood

The tech stack hits the sweet spot between accessible and impressive. Three.js handles the graphics layer with custom low-poly cubic models, orbit controls, and a localized 1400K bonfire flame illumination matrix that actually looks atmospheric rather than like a Unity tutorial project gone wrong. On the AI side, Yeole integrated Google AI Studio SDK with the gemini-2.5-flash model for multi-turn, state-retaining dialogue pipelines. The real craft is in the prompt engineering—each voxel entity gets isolated contextual instruction blocks so they don't bleed responses into each other.

Prompt Engineering as Gameplay Mechanic

Here's where it gets clever: Yeole designed the AI to occasionally slip up and use "technical, precise engineering metrics or mathematical jargon" before self-correcting to sound natural. The system computes an AI Certainty Meter that scales dynamically based on structural linguistic patterns exposed during play. So you're not just chatting—you're analyzing discourse architecture in real-time. One of Yeole's sample configurations shows the rogue agent persona assigned as "Sleek Billy, an experimental AI entity operating within a blocky physical chassis" with explicit instructions to simulate organic conversation while triggering subtle context slips.

Hackathon Credentials

The project is submitted for evaluation under two prize categories: Best Ode to Alan Turing and Best Google AI Usage. The former earns points for implementing the classic test with an interactive 3D avatar of Turing himself as Simulation Director overseeing the diagnostic boundary. The latter demonstrates programmatic control through strict schema compliance, system instruction isolation, and dynamic contextual state clearing during runtime cycles—solid execution on Google's SDK.

Key Takeaways

  • Built with Three.js (low-poly voxel aesthetics) + Google AI Studio SDK (gemini-2.5-flash)
  • 10-turn constraint creates urgency; AI Certainty Meter adds scoring tension
  • Prompt engineering is the gameplay: watch for technical slips and conversational repairs
  • Deployed live on Netlify with full source on GitHub

The Bottom Line

This is exactly the kind of project that makes me love this community—taking a foundational concept from computing history, wrapping it in something visually distinctive, and shipping it so others can poke at it. Whether you think you'll ace it or get fooled by Sleek Billy's charm offensive, Solstice Turing Simulation is worth bookmarking. The live demo and GitHub repo are linked below if you want to run the experiment yourself.