The workshop door opened just wide enough for people to slip through single-file. Ryan went first, then the two professors, then Tom and Lisa bringing up the rear.
Ward stepped inside and stopped.
He'd seen the video. He'd watched it eleven times. He knew the dimensions, had estimated the mass, had run napkin calculations on the power requirements. None of that prepared him for the physical reality of standing at the feet of a forty-foot steel skeleton.
Scrapper filled the workshop the way a cathedral fills its nave not just occupying the space but defining it. Everything else in the building existed in relation to the mech. The tools, the workbenches, the generator, the cracked concrete all of it was context. Scrapper was the sentence.
"Well," Hartley said quietly. "That's bigger than I expected."
Ward was already moving. He crossed to Scrapper's nearest foot and placed his palm flat against the ankle joint. Cool steel. He rapped it with his knuckles. Tapped the weld seam where the ankle housing met the lower leg strut. Ran his fingers along the grain of the metal.
Plain carbon steel. Maybe A36, maybe something slightly better. Nothing exotic. Nothing that would raise an eyebrow at any fabrication shop in the country.
Which made the video even more confusing.
"Can we see it operate?" Ward asked, turning to Ryan.
Ryan shook his head. "Not today. The diagnostic report flagged material degradation across several subsystems most of the components are running on borrowed time. I'm saving the remaining operational cycles for tomorrow's livestream. But I can show you the diagnostics, walk you through the engineering, answer any questions."
Ward nodded. Fair enough. If the kid was telling the truth about component degradation, wasting a cycle on a private demo would be stupid. If he was lying, the livestream tomorrow would expose him anyway.
"Then let's start with the obvious question," Ward said. "How do you control it? In the video, there's no visible control interface beyond what looks like a pair of gloves and some foot pedals. No joystick arrays, no hydraulic feedback rigs, nothing consistent with operating a system this size. What am I missing?"
Ryan hesitated. Ward caught it a pause just long enough to mean the kid was deciding how much to say, not stalling for time.
"I built a system I call a neural link," Ryan said. "It connects the pilot's nervous system directly to Scrapper's control architecture. Motor signals from the brain get translated into machine commands in real time hand movements, weight shifts, balance corrections, all of it. The pilot thinks, the mech moves."
Silence.
Ward glanced at Hartley. Hartley's expression hadn't changed, but Ward had known the man for fifteen years and could read the micro-shift a slight tightening around the eyes, the jaw setting half a degree. Translation: this kid just lost me.
Neural-mechanical interfaces were a real field. DARPA had programs. Universities had programs. The best of them could manage a cursor on a screen or a robotic arm performing gross motor tasks with a two-second delay. What Ryan was describing real-time full-body motor translation with the fidelity to balance a thirty-ton bipedal system — wasn't just beyond current capability. It was beyond current theory.
It was like a street vendor hanging a sign that read: PROFESSIONAL AIRCRAFT CARRIER REPAIR — ALSO NUCLEAR WARHEAD MAINTENANCE — $5.
Ward kept his face neutral. "That's an extraordinary claim."
"I know how it sounds."
"You're describing a technology that would represent, conservatively, a thirty-year leap over the current state of the art in brain-machine interfaces."
"Yes."
"Built by a fourteen-year-old."
"Also yes."
Ward let it sit. There was no productive way to challenge the claim without seeing it operate, and the kid had already explained why he wasn't demonstrating today. So Ward did what any good scientist would do when presented with an extraordinary claim and insufficient data: he set it aside and looked at what he could evaluate.
"Walk me through the rest of it," he said.
Ryan did.
They spent the next hour circling Scrapper at ground level while Ryan pointed out design details that weren't visible in the video. Joint articulation geometry. Load distribution pathways. The cable routing that ran through Scrapper's torso like a nervous system of its own. The way the hip joints used a compound pivot that allowed rotational movement in three axes simultaneously.
Ward listened with increasing discomfort.
Not because the explanations were bad. Because they were good. Too good. The compound hip pivot alone was an elegant solution to a problem that most robotics engineers solved with bulky multi-joint assemblies. The cable routing minimized signal loss while maintaining flexibility across the full range of motion. The load distribution wasn't textbook — it was better than textbook, accounting for dynamic stress patterns that only mattered in bipedal systems, which almost nobody built at this scale.
These weren't the answers of a kid reciting a script. These were the answers of someone who understood the engineering at a level that Ward found frankly unsettling.
Hartley, who'd been quiet for most of the tour, finally spoke up at the base of Scrapper's spine.
"This torso junction," he said, pointing to a cluster of structural members where the upper and lower body met. "The load path splits into four channels here and reconverges at the hip mounts. That's not a standard truss pattern. Where did you get this design?"
"I came up with it," Ryan said. "Standard truss patterns don't account for the lateral forces generated during bipedal locomotion. If you use a conventional X-brace here, the torso twists under asymmetric leg loading. The four-channel split distributes the torsional stress evenly across both hip mounts."
Hartley stared at the junction for a long time. Then he looked at Ward.
Ward looked back.
Neither of them said anything.
Ryan led them up the ladder next — all the way to the cockpit, thirty feet above the cracked concrete. The two professors climbed without complaint, though Hartley's grip on the rungs was noticeably tighter than Ward's.
The cockpit was as bare as the video suggested. Seat. Vest. Pedals. Straps. Touchscreen. Gloves. No armor. No canopy. Open to the air on three sides. The wind at this height was noticeable.
Ryan powered up the touchscreen — battery backup, independent of the main generator — and ran the self-diagnostic suite. Lines of data cascaded across the display. System statuses. Component health ratings. Thermal histories. Stress accumulation logs.
Ward leaned in. The diagnostic interface was clean, well-organized, and responsive. The data looked real — not the kind of thing you'd mock up for a demonstration, because nobody would bother faking thermal gradient logs for individual actuator mounts.
"This diagnostic suite," Ward said. "You wrote this?"
"From scratch."
"What language?"
"Custom. Based on C++, but I needed real-time interrupt handling that the standard libraries couldn't give me, so I wrote my own kernel-level interface."
Hartley made a sound in the back of his throat that might have been a cough and might have been his worldview cracking.
They climbed down. Said their goodbyes. Shook hands with Tom and Lisa. Walked out through the gap in the workshop door and back into the Texas afternoon.
They didn't speak until they were in the rental car with the doors closed and the AC running.
"So," Ward said.
"So," Hartley agreed.
A long pause.
"Is it real?" Ward asked.
Hartley rubbed his face with both hands. "The neural link claim is insane. Absolutely insane. If anyone else on Earth told me they'd built a real-time full-body neural-mechanical interface in a backyard workshop, I'd walk out of the room."
"But?"
"But that torso junction is one of the most elegant structural solutions I've seen in twenty years. And he didn't hesitate when I asked about it. He didn't recite an answer — he reasoned through it, in real time, like he'd thought about the problem from six different angles before I even opened my mouth."
Ward nodded slowly. "The diagnostic suite bothered me. The thermal logs are granular down to individual actuator mounts. Nobody fakes that level of detail for a hoax. There's no audience for actuator thermal data. You'd only generate it if you actually needed it."
"So what are we looking at?"
Ward stared out the windshield at the flat Texas horizon.
"Even if this is a fake," he said carefully, "it's the most sophisticated fake I've ever encountered. The physical engineering alone — the frame, the joints, the load paths — represents years of serious work by someone who understands structural dynamics at a graduate level. At minimum."
"He's fourteen, Doug."
"I know."
"He wrote his own kernel-level interface."
"I know."
Hartley shook his head. "What do you want to do?"
"Come back tomorrow. Watch the livestream in person. If that mech moves the way it did in the video — with that control fidelity, on one power feed — then we're not looking at a hoax. We're looking at something that needs to go up the chain immediately."
"Up the chain meaning—"
"Meaning I make a phone call to Lincoln Lab and tell them to turn on their television."
Hartley sat with that for a moment. Then he started the car.
"You're buying dinner," he said.
"I thought that was only if it turned out to be a science fair project."
"It's turning out to be a headache, which is worse. You're buying dinner."
Ward didn't argue. He was already composing the email in his head.