The black wave broke across the horizon and kept coming.

Shen Cong felt his face go tight. The first hundred workers he could manage — one at a time through the window gap, Vajra’s Active armor handling anything that reached the hull. Against a tide like this, that approach became meaningless. The hull would be covered before he could make a dent in the numbers.

He sealed the window panel, took one deep breath, and made himself think.

The apocalypse kept producing threats his planning hadn’t accounted for. The flood hadn’t been in his scenarios. This wasn’t either. What they had in common was that Vajra’s design, however carefully considered, had no clean answer for either of them. He’d already started working on the waterproofing problem after the flood — sealing the connection joints, eventually adding motor-driven propellers for underwater movement. He hadn’t started working on the ant problem because he hadn’t known the ant problem existed.

He thought briefly about fire and water — the standard answers to ant swarms. No water source nearby. No significant combustibles in range. The terrain prevented him from simply driving away at speed.

He went to the weapon rack and started pulling hardware.

High-pressure air rifles — three of them, self-loaded lead pellets, all custom-built by a seller he’d contacted before the apocalypse, claimed accurate and lethal against dogs at a hundred meters. He’d verified that claim against smaller targets and believed it. He opened the window facing the approaching swarm and started shooting.

The larger ants were clearly soldier variants — Coca-Cola-bottle scale, mandibles oversized even relative to their enlarged bodies, moving with a purposefulness the workers hadn’t shown. He prioritized them.

Crack. Crack. Crack.

Each shot dropped one. Each dropped one was immediately irrelevant against the density of what was coming. He kept shooting anyway because doing something was better than doing nothing, and because the math on the gas cylinder meant he had approximately fifty shots at full power before the pressure dropped and the terminal velocity degraded.

He counted. Fifty shots fired. Pressure dropping. The swarm had reached Vajra’s perimeter.

He switched to the fang-knife, Activity-infused, working the window gap with the same controlled-engagement approach as before — plate in to block, dispatch what got through, plate out, repeat. The technique worked. The problem was the cost.

Activity-enhanced cutting was necessary because the ants had developed hardened exoskeletons — the same metallic quality he’d seen in the Activity cores of other mutated creatures, distributed across the whole exterior surface rather than concentrated in one location. Without the Activity sharpening, the knife dragged. With it, each swing consumed a measurable amount of his reserves.

He kept cutting. His forearm started to burn. Sweat ran into his eye and he didn’t have a hand free to wipe it.

More every wave. Not slowing down.

He was good at this. He’d always been good at mechanical problems, at finding the angle that made an impossible thing possible. But good at mechanical problems didn’t mean good under pressure, and he knew that about himself. The X-ray incident with the crocodile had been a mechanical-problem-solver making a pressure decision — I need to know what this is — without properly weighing the cost of knowing it. The crocodile had nearly destroyed Vajra.

He’d made a framework from that lesson. Simple, blunt, practical:

Can beat it: don’t let it go. Can’t beat it: run immediately. Can’t run: play dead. Playing dead fails: fight to the end.

He closed the window and pressed his back against the hull, breathing hard, working his wrist back and forth to get feeling into it.

Playing dead. How do you play dead against ants?

He focused his mind and extended into Vajra’s systems.

The feedback was immediate and bad.

The ants were on the hull. All of it. Every surface covered. They’d found no entry point and had begun working on the armor plating directly — but the Active armor was resisting. The hardness and density the Activity provided made the metal effectively impenetrable to their mandibles.

Then the venom reading came through.

The ants were secreting something from their stingers — a substance that worked like formic acid, but wasn’t. It wasn’t attacking the metal directly. It was attacking the Activity in the metal. Corroding it. Dissolving it at a rate he could measure in real time.

49% saturation. 48.7%. 48.4%.

One minute: 0.3% lost.

He ran the math. From 49% to the 10% survival threshold was 39 percentage points. At 0.3% per minute, that was approximately 130 minutes before Vajra entered critical failure.

Just over two hours. Then the armor lost its Active resistance. Then it became ordinary steel. And ordinary steel, he’d watched from a distance, dissolved in minutes under a full swarm.

He started pulling in ambient Activity as fast as he could manage, compensating for the drain. The rate of loss slowed fractionally.

Not enough.

He tried going still — cutting all the noise and movement from Vajra’s systems, hoping the ants would lose interest. Nothing changed. They’d located something worth attacking and they weren’t built to reconsider.

Different problem than the crocodile. Silence doesn’t work here.

He started going through his knowledge of the Activity system — every property he’d identified over the past two and a half months, laid out in sequence:

Extension. Amalgamation. Conduction — Activity propagated well through metal, slowly through non-metal. Exchange. Absorption. Devour. Sharpening. Recovery. Reinforcement. Radiation. Frequency. Intake. Mutation.

He got to frequency and stopped.

Different mutated organisms radiated Activity at different frequencies. He’d noticed this when absorbing the motorcycle’s Activity — the friction of two different signatures running through the same system before the digestion process aligned them.

He knew how humans and animals distinguished one thing from another: visually, primarily, with smell and sound as support systems. He’d read enough about ants to know they operated on touch, smell, and pheromone signaling, with their antennae doing most of the work. Disrupt those sensing channels and you disrupted their ability to identify targets.

Mutated creatures had an additional channel: Activity perception. They felt the radiation from other Active objects. Frequency was how they distinguished between them.

Sound: handled. Vajra’s engine is off.

Smell: need a strong masking agent.

He knew what he had for that.

Benzene hexachloride — BHC, technically hexachlorocyclohexane, banned from commercial production decades ago due to persistence and toxicity, one of the most effective broad-spectrum insecticides ever formulated. He’d sourced roughly ten kilograms of it through the same unofficial channels he’d used for the air rifles, packed and sealed in the supply section. He’d reasoned that chemical pest control might matter in a post-apocalypse environment. He hadn’t been wrong.

He found it, opened the window a crack, and distributed a heavy dusting across the ground surface and lower hull sections.

The effect was immediate. Ants in contact with the powder retreated. A clear zone formed around the application areas, spreading outward.

That handled smell. Now the frequency problem.

The ants were sensitive to Activity radiation. They’d found Vajra because of it. If he could change Vajra’s frequency signature — match it to something the ants didn’t recognize as a target, or better yet, match it to something they recognized as the same as themselves —

He had a live ant colony crawling on his hull.

He knew what frequency they were broadcasting.

(End of Chapter 23)