He Built a Self-Driving Car in His Garage in 30 Days.

BUILDERS SERIES · ISSUE 003

George Hotz — known online as geohot — never applied for a job at the companies he competed with. He just built things that proved he didn't need to.

The Kid in New Jersey Who Couldn't Leave Things Locked

George Hotz grew up in Glen Rock, New Jersey. His father ran technology for a high school. His mother was a therapist. By 14, he was a finalist in the Intel International Science and Engineering Fair for building a robot that could scan a room and map its dimensions. He appeared on the Today Show and Larry King for it. He was in eighth grade.

Long before anyone was paying attention, he was solving problems that most adults wouldn't attempt — not for a grade, not for a company, because the problem was there and unsolved.

The hacking came from the same place. Not malice. Curiosity. A refusal to accept that something was locked if he could find a way through.

The iPhone. Two Months After Launch. Alone.

In the summer of 2007, Apple released the original iPhone. It was tied to AT&T. In August 2007, seventeen-year-old George Hotz became the first person to remove the SIM lock — making it the first reported iPhone jailbreak in history.

He did it in his bedroom. No team. No funding. No one had done it before — not researchers, not rival carriers, not the global community of professional security engineers who had been trying. A teenager in New Jersey got there first.

After the jailbreak, Apple co-founder Steve Wozniak sent him a letter of congratulations.

He was 17. He'd never finished high school.

Then the PlayStation 3.

The PS3 was considered one of the most secure consumer devices ever made. Sony had engineered it specifically to resist exactly the kind of attack Hotz had pulled on the iPhone. The security community largely viewed it as uncrackable.

He became the first person to run through the hard-core defence systems in the Sony PlayStation 3 and crack it open too.

Sony's response was to sue him. The lawsuit became a cause in the hacker community — the image of a corporation deploying its legal department against a teenager in his bedroom for the crime of understanding how their product worked. They reached an out-of-court settlement, on the condition that Hotz would never again resume any hacking work on Sony products.

He moved on. There was always something else to break open.

The Detour. And the Decision.

He went to Rochester Institute of Technology but dropped out after one semester. He interned at Google, worked at Facebook, took computer science courses at Carnegie Mellon but didn't earn a bachelor's degree. At every institution, the same pattern: arrives, demonstrates he's operating above the level of the curriculum, leaves.

He was, in his own words, on a walkabout — trying to work out what deserved his full attention. The iPhone was done. The PS3 was done. He needed a problem that was genuinely unsolved at a scale that mattered.

He found it in self-driving cars.

The Garage. Thirty Days. One Car.

In late 2015, a journalist arrived at Hotz's house in San Francisco to find a white 2016 Acura ILX in his garage, fitted with a lidar system on the roof, a camera near the rearview mirror, and a tangle of electronics where the glove compartment used to be.

Hotz had built it in about a month.

The claim was, on its face, absurd. Google had been working on self-driving cars for years with hundreds of engineers and hundreds of millions of dollars. Tesla was deploying Autopilot across its entire fleet. Waymo had logged millions of miles of testing. The conventional wisdom was that autonomous driving was a decade-long engineering challenge requiring vast teams, vast capital, and vast infrastructure.

Hotz had done it in thirty days in his garage. Then he drove it onto a California highway and let a journalist watch.

comma.ai.

And a Different Theory of the Problem.

He founded comma.ai in September 2015. The idea was radical in its simplicity: self-driving wasn't a hardware problem requiring expensive sensors — it was a data and software problem, and the data was already being generated by millions of ordinary drivers every single day.

While bigger companies were developing complex systems with expensive LIDAR and other sensors, Hotz was trying to bring plug-and-play driverless technology to the masses — running it on a phone, with a tiny team and minimal funding.

His team crowdsourced driving data from real users on real roads in real cars, fed it into neural networks, and let the system learn by watching how humans actually drive. No hand-coded rules. No bespoke sensor arrays. No fleet of specially equipped test vehicles. Just data and a model that learned from it.

When regulators sent a cease-and-desist letter threatening daily fines, Hotz scrapped the hardware product and released his self-driving software, openpilot, as open source — along with open-source plans for the hardware platform. If he couldn't sell it, he'd give it away. Within four years, Consumer Reports ranked openpilot above all other driver assistance systems, including Tesla Autopilot, Cadillac Super Cruise, and Ford Co-Pilot 360.

There Are Builders Like This Everywhere. Most Have No Platform to Show It.

Hotz got funded because a Bloomberg journalist happened to show up. Because a16z's Chris Dixon tested the car in person. Because the iPhone hack was public enough, and loud enough, that the right people already knew the name geohot before he walked in the room.

His work was visible. But only because the circumstances made it visible.

Most builders working at that level don't get a Bloomberg feature. They don't have a famous hack on their record that precedes them into every room. They're building things just as audacious — in garages, in bedrooms, in evenings after day jobs — with no event that forces the right people to pay attention.

The work exists. The proof is there. The companies that need them just have no way to find it.

That's exactly what Vaultt is built for. Not a CV. Not a list of job titles and qualifications. A place where the work is the profile — where a builder can show what they've actually made, and the companies that need that kind of mind can find it directly.

Geohot didn't need Vaultt. His work announced itself. But for every geohot whose work breaks through, there are hundreds whose work is just as real, just as proven — and completely invisible to the people who'd hire them on the spot if they could only see it.