Building 'Short Circuit' in 6 Hours: A Lesson in AI Orchestration

Sometimes, the best constraints are the ones forced upon you by the clock. For the 2026 GameDev.tv Game Jam, my availability was entirely compressed. With only about six hours to spare from inception to deployment, I set out to build Short Circuit—a compact, arcade-style 3D spatial puzzle wrapped in a hardware terminal aesthetic.

Beyond just shipping a working game under a brutal deadline, I had a secondary technical objective: to test the limits of AI orchestration. I wanted to see how efficiently I could leverage an AI collaborator as a rapid-fire sounding board and code compiler to compress days of typical development layout into a single afternoon sprint.

The game itself is a mechanical twist on the classic Snake. Instead of chasing pellets to grow longer, you are routing a critical signal line across a compact 10x10 component matrix, fighting against a highly restrictive, dynamically calculated wire spool budget.

Live in-browser prototype embedded below (Use Arrow Keys to launch your connection route and navigate; press R to reboot hardware on failure):

The AI Co-Piloting Workflow

Orchestrating an AI to build code under a tight deadline requires a shift in mindset. You can’t just treat it like a search engine; you have to treat it like a junior engineer pairing with a systems architect.

My strategy was to act as the core architect—defining the geometric boundaries, structural logic gates, and game-state rules—while using the AI to rapidly compile the unified GDScript files. When paste errors or syntax discrepancies popped up (such as a mashed variable initialization block deep in the instantiation code), the loop relied on immediate diagnostic feedback to hot-fix the pipeline.

By treating the AI as an active engine component, we managed to cleanly integrate edge-spawning normal rotations, synchronous startup highlighting, and paused-input emergency loops in a fraction of the time it would take to reference documentation manually.

The Core Architecture

With the pipeline established, we focused heavily on ensuring the gameplay mechanics felt tight, predictable, and fair.

1. Dynamic Spool Calibration Math

Because the start position of the player and the target edge outlet are completely randomized at startup, a hardcoded wire budget would make the game mathematically impossible on some seeds and trivial on others.

To solve this, the engine calculates the straight line-of-sight distance between your starting component hub and the destination terminal on frame one using a standard distance formula. The system takes the square root of the combined horizontal and vertical differences squared, and then applies a variable multiplier buffer. This scales your total wire resource allocation up by 40% to 80% dynamically, ensuring every single map layout is a bespoke riddle with a razor-thin margin for error.

2. Streamlining the UX: Directional Launching

When dealing with sudden grid movements, clicking an overlay button to start feels clunky. I implemented a directional arrow key launch gate. The game completely freezes state at initialization while the source and target nodes pulse in sync, signaling the connection points. Striking an input direction immediately injects velocity along that chosen vector on frame one, eliminating clunky startup lag.

3. Overcoming the Paused-Input Trap

One of the trickiest bugs encountered during the sprint involved Godot’s pause tree logic. When a collision or short-circuit occurred, calling a hard pause instantly froze the entire processing engine—meaning the player’s emergency “R” key reboot inputs were being completely ignored. By forcing the script node’s process mode parameter into an unconditional “Always” state via code, the hardware reboot sequences remain fully operational even when the physics simulation is completely halted.

Takeaways

Is Short Circuit a flawless masterpiece? Far from it. Under a 6-hour timeline, visual flourishes like background textures and advanced shaders had to be cut to protect the core deadline.

However, as a pure exercise in rapid prototyping and AI workflow management, it’s a project I’m glad to put my name on. It proved that when properly orchestrated, human-AI collaboration can turn a compressed timeframe into a fully functional, deployed reality.

The project is fully live and playable above. If you manage to establish a connection with a significant spool surplus, let me know your high score!