2026 IEEE Metaverse Grand Challenge for Simulation-Based Learning
Most circuit simulators check your wiring against an answer key and tell you that you are wrong. CircuitCoach reports what a real instrument would have shown you, then puts an AI tutor on the specific mistake you actually made.
A real breadboard, a component kit and a bench multimeter cost real money. The students who would benefit most from a bench are usually the ones who never get one.
And a simulator that only says incorrect teaches the wrong lesson. On real hardware nothing announces your mistake. You get a reading, and you have to work out what the reading means.
It runs in a browser with no install and no hardware cost. Holes are wired like real copper, so five holes in a column are one node and the top rails are not the bottom rails.
Every fault is a measurement. Reverse an LED and you get
V_LED = −5.00 V, zero current, junction blocking. Never
try flipping the LED.
Seven minutes, narrated, recorded against the live application. The AI hints in this take are real generated output, not staged text.
Streamed from Drive, so nothing heavy ships with the page. If the frame is blocked, open it directly. The voiceover transcript is published with timecodes.
Difficulty moves from wiring, to arithmetic, to reasoning about logic. The board never changes.
Tier 1
Build a closed loop, and get the polarity right. The LED body sits off centre so the anode reads as the long leg, the same cue you use on a real part. Twelve distinct faults are reachable here, from a hard rail short to a parallel path that quietly bypasses the emitter.
reversed_polarity · V_LED = −5.00 V · I = 0.00 mA
Tier 2
Same circuit, but the resistor value is the answer. You pick from real E12 stock, 100 to 470 ohms, rather than typing a number. Undersize it and the LED runs white hot, then dies past twice its rating. Oversize it and it lights visibly dim. Both directions fall out of Ohm's law.
target band 65 to 100% of rated drive
Tier 3
A logic block with an input pin and a fixed sketch: output follows input. A button with no pull resistor leaves the pin floating, and that is its own fault with its own reading. Because the sketch never changes, the opposite condition needs the opposite wiring, so it has to be reasoned through rather than copied.
input_floating · R to SRC+ = open · R to SRC− = open
The escalation decision is deterministic and runs in the browser. The model is used only to word the hint, so the teaching logic cannot misbehave.
Attempt 1
Visual feedback only. The board shakes, the LED stays dark, no hint.
Attempt 2
A question that points at the concept without handing over the fix.
Attempt 3
Names the specific component, terminal or value involved.
Attempt 4
States the fix, then explains in one sentence why it works.
The things worth weighing in an evaluation.
Holes are unioned into nets, then every simple path from supply to return is enumerated and classified. Forward biased, reverse biased, or bypassing the emitter entirely. No string matching against an expected answer.
86 assertions on the solver in plain Node, and 112 driving the real interface with real DOM events in headless Chrome. The tutor suite runs against a stub so it never spends API quota.
The whole circuit can be built from the keyboard. Arrow keys walk the hole grid, letters arm parts, space cycles power. It opens from a file with no server and no build step.
The API key stays server side and never reaches the browser. Static serving is restricted to the front end, request bodies are capped, and every field is validated before anything reaches the model.
Seventeen named codes, each with its own instrument reading. This is also the input vocabulary the tutor reasons over.
No account, no install. Build the loop, get the polarity wrong on purpose, and read what the bench tells you.