0.11 seconds.

That was the "Danger Sense" activation threshold. The move was executed before his conscious brain had even finished receiving the sensory input. It was pure, raw instinct, processed by a nervous system that had been rebuilt by a million virtual laps.

GD-02 had no response. Its dataset was a library of every lap ever driven by a human, but it didn’t have a record for a line that was invented 0.11 seconds before it was driven.

Leo felt the car pivot. It was a violent, sharp rotation, the rear of the Arcadia car stepping out just enough to point the nose at the exit. He didn’t counter-steer; he balanced the slide with the throttle, his foot dancing on the pedal with micro-adjustments.

He shot through Sainte Dévote and began the climb into Massenet. He checked the data feed.

The gap said 0.0.

The cars were level.

---

[SIMEX BROADCAST, LAP 22 of 30 (92) SECTOR 1:]

[SECTOR 1: LEO KAITO, 27.4 seconds. New session record.]

[GAP TO GD-02 AT MASSENET: 0.0 SECONDS. THE CARS ARE LEVEL.]

[GD-02 note: The precise profile does not have a response model for a car running level at Massenet. The optimal line through Casino Square assumes clear air ahead. With a car at equal position, the aerodynamic condition is altered. GD-02 must choose: defend or maintain optimal line.]

[GD-02 has chosen to maintain optimal line.]

[This is a rational decision. It is also the decision LEO KAITO anticipated.]

---

Leo could feel the AI’s logic working. GD-02 was the "Professor." It was built on the idea that the fastest way around the track was the only way that mattered. It believed in the "Optimal Line" like a religion.

When faced with a human driver running side-by-side at 180 kilometers per hour, the AI didn’t panic. It didn’t try to "squeeze" Leo into the wall. It simply calculated that defending the position would cost it 0.2 seconds in sector time.

Therefore, it stayed on its optimal path, trusting that its superior exit speed would eventually win out.

"You’re too smart for your own good," Leo hissed, his teeth gritted.

They reached Casino Square.

Side by side. It was a terrifying sight. Two cars navigating a narrow, wet street with less than a hand’s width between them. In the real world, this would be a death wish. In the simulation, it was a battle of mathematical certainty.

GD-02 took the inside. It was the correct choice. The inside line was shorter, allowed for an earlier apex, and provided a straighter shot to the next corner. It was the line that won races.

Leo took the outside.

In isolation, the outside line was a failure. It was longer. It was wetter. It required more steering input. Any coach in the world would have told Leo he had just lost the position.

But Leo wasn’t driving in isolation.

He had built this outside line three corners ago. The specific load he had carried out of Massenet, the way he had shifted the car’s weight to the right-side suspension, meant that his tires were in a unique state of compression.

He wasn’t fighting the car to stay on the outside; the car was leaning into the turn, using the centrifugal force to "hook" onto the tarmac.

The outside line wasn’t a compromise. It was the final word in a sentence that had started at the first corner.

GD-02 came out of Casino Square 0.2 seconds ahead. Its optimal line had given it the faster exit, just as the math predicted. The silver car surged forward, its rain light blinking triumphantly.

But Leo didn’t lift. He felt the front-left tire bite. Because he had taken the wider arc, he had kept the car’s momentum higher. He wasn’t accelerating from a lower speed; he was maintaining a higher one.

Through Mirabeau, the gap shrank to 0.1 seconds.

Into the tunnel, the gap was 0.05.

The tunnel was a tunnel of white noise and strobe lights. Leo could feel the heat radiating from GD-02’s exhaust. The "Slipstream Prediction" lines were a solid wall of purple light, pulling him forward like a magnet. He didn’t even look at the track. He looked at the air. He followed the flow-lines, weaving through the turbulence.

They came out of the tunnel’s mouth, the grey daylight blinding for a fraction of a second.

They were level.

Leo didn’t wait for the braking zone. He didn’t wait for the AI to move. He committed to the Nouvelle Chicane first apex before the corner even appeared. He threw the car into the gap, a space that existed only in his mind and the telemetry.

He went through GD-02 like a ghost.

The AI, programmed to avoid contact at all costs, had to yield. Its logic core couldn’t find a way to hold the line without a 92% chance of a terminal crash. For a machine, 92% was a certainty.

It backed off.

Leo shot through the chicane, the car dancing on the edge of a spin, and pulled away.

---

[SIMEX BROADCAST, LAP 22 of 30 (92) REVIEW:]

[LAP 22 OF 30. RESULT: LEO KAITO, CLASSIFIED P1.]

[Lap time: 1 minute 10.6 seconds. Session record. Margin over previous session record: 0.3 seconds.]

[The decisive sequence began at Sainte Dévote and concluded at the Nouvelle Chicane; a full-lap construction that used the tire state, crosswind data, and GD-02’s predicted line choice as inputs across seven corners, producing an overtake opportunity that did not exist as a discrete decision at any single point in the lap.]

[The human driver did not decide to overtake GD-02 at the Nouvelle Chicane. The human driver decided nothing. The overtake was the output of a system that had been running since the standing start.]

[GD-02 analysis: The precise profile’s optimal line was correct at every individual corner. The aggregate of individually correct decisions produced a suboptimal outcome. GD-02 was outdriven not at any corner but across all of them simultaneously.]

[This is not a modeled failure mode for GD-02.]

[Points awarded, LAP 22 of 30 (92):]

[1st: LEO KAITO, 10 points]

[2nd: GD-02, 7 points]

[GRID RANKINGS, LAP 22 of 30 (92):]

[1st: GD-02, 147 points]

[2nd: LEO KAITO, 122 points]

[Gap: 25 points. Eight laps remaining.]