On the morning of Friday, October 22, 2022, Rahul Sonwalkar was at the gym in San Francisco. It was a cool, cloudy day, and just around the corner, a clique of journalists and camera crews had assembled at the doors of 1355 Market St, otherwise known as “The Twitter Building.” That Wednesday, Elon Musk walked through those doors carrying a sink. “Let that sink in!” the new owner tweeted.
But the best, most elaborate play on words that week did not belong to the boss. It’d belong to one of his (fake, not real) employees: Rahul Ligma, an invention of Rahul Sonwalkar. That Friday morning, Sonwalkar noticed the cameras when he walked into the gym. An hour later, they were still there, still trained at the doors of Twitter, waiting for something, anything.
Sonwalkar, wondering why the press didn’t have anything better to do, texted a friend of his, asking whether he’d want to pull a joke on the cameras: pretend to be employees that were sacked by Elon Musk leaving the building with their belongings. The friend, Daniel, brought two boxes. Each had a copy of Michelle Obama’s memoir, “Becoming.” They turned the corner toward the cameras, walking with long faces. For the first time that morning, the cameras moved abruptly.
They explained their situation, and gave their names: Rahul Ligma and Daniel Johnson. The joke went straight over the journalists’ heads, and into their stories. Ligma and Johnson had been callously fired by Elon Musk. Mr. Johnson lamented that he’d have to explain this to his “wife’s boyfriend” and needed to get home. Mr. Ligma complained that he wouldn’t be able to afford to charge his Tesla.
Reporters gave them business cards so they could get in touch with the full story of their unceremonious firing.
Only later did they start to catch on. “They started blowing up my phone,” Sonwalkar told me. “We didn’t expect it to go viral like it did, we just thought maybe some of our friends would see it and find it funny.” When word of the joke got out, it became an immediate hit. Two practical jokers had exposed the media bias around Musk’s purchase of Twitter — bias so strong that they’d report anything from anyone. The Watergate source who took down Nixon was Deep Throat; the Twittergate source that took down the press was “Ligma Johnson.”
A few weeks later, Musk invited the two to Twitter HQ for a “reunion” to apologize for “firing” them – and to celebrate the gag they’d pulled on the august press.
On November 22, 2022, the Wikipedia entry for “Rahul Ligma” was published.
“A lot of things can be engineered. Social engineering, it’s just like computer engineering. It’s not that different, because people are systems with vulnerabilities,” Sonwalkar told me when we met for this piece. We met at his office in the heart of San Francisco’s new AI scene.
He’s the founder and CEO of Julius AI — a computational AI platform that specializes in writing code in Python and R. Today, about a year since public launch, Julius has over a million users.
It’s been a whirlwind four years for Sonwalkar, who in 2020 was finishing a computer science degree at a school that hasn’t played a starring role in the technology world thus far: The University of Texas at Dallas, which awarded him a scholarship. The biggest tech companies didn’t recruit at UTD. So, Sonwalkar found his way to Silicon Valley through another avenue: hackathons.
“Hackathons were really where I found my community,” he told me. “I met people who were building things just because they could.” At hackathons from coast to coast as a new American, Sonwalkar met friends who referred him for internships. He worked at Facebook (the subject of the work was, amazingly, disinformation) and Uber.
Sonwalkar participated in the Summer 2022 session of Y Combinator, the startup accelerator, and since then has been working full time on Julius. In the year since Julius launched, he’s personally answered 5,000 support tickets for users. It keeps him close to the product, and to the customers. One source of unexpected popularity? Scientists, for whom Julius is a powerful tool.
Sonwalkar and his team did not expect the popularity in the scientific community. “But they love it,” he told me. “And they’re spreading the word and talking about it with everyone.” It turned out that unlocking advanced coding for academic scientists saved them time and made results richer. Sixty years ago, scientists had to do calculations and regressions on bulky computers running FORTRAN, with IBM punch cards for data input and output (above). Professionals ran the computers. Julius gives scientists the power of code at their fingertips. The day before we met, an ophthalmology professor in Germany had sent a testimonial explaining the ways he used Julius in his research.
It’s gratifying for Sonwalkar, who counts many physicians and medical researchers among his users. “If we can speed up critical research by even .001%, the company will be worth it,” he said. But course, his plan is much bigger than marginal improvement.
“We want to become the biggest company ever”
Imagine this: for every one person in a business that can do science and data analysis, there’s someone in finance, sales, operations, supply chain, maybe even HR or legal, who has a massive amount of data — but the expertise to analyze it.
“Code is one of the most powerful tools invented in the computer age,” Sonwalkar told me. “It can move markets, create software, turn ideas into reality, and transform massive amounts of information into actionable insights.” Yet almost nobody on Earth knows how to write code, how to use this superpower.
With the advent of AI models useful for coding, Sonwalkar sees two potential dominant paths. The first is that LLMs will just make existing programmers more productive, creating super engineers with more output than ever. The second version, which Sonwalkar and Julius are betting on, is that AI will be like a programmer in everyone’s pocket. A billion or more people could be able to code without formal skills.
Knowing this, it’s no surprise that he says, “We want to become the biggest company ever.” As Sonwalkar explains it, today one has to buy software, learn how to use it, and pay for it every month to keep using it. He envisions a world in which the people facing the problem — doctors, accountants, scientists, basketball coaches — can build software for themselves, totally custom. He compares it to clay, which each person is free to mold to his or her needs. “And I think that it is a bigger opportunity than the market capitalization of all software companies today.”
At the beginning of the computer age, every person with a computer was a programmer, because they had to be. They had to program the computers before user interfaces came along. It struck me that Sonwalkar, in a way, wants to get us back to that, back to when everyone could program. Except this time it won’t be just born tinkerers and hackers. It can be everyone. A billion people.
We’re not quite there yet, as he’s first to admit. But AI models are improving rapidly. As models improve, particularly with the advent of more advanced hardware like Nvidia’s B100 Blackwell architecture (coming later this year or early next) the capabilities of AI programmers will expand dramatically. The market, in his view, hasn’t yet priced in the impact of these coming advancements.
And considering Julius has been available for just a year, exciting times ahead are guaranteed.
“I think San Francisco is weird again,” Sonwalkar told me. And to be clear: that’s a good thing! “One thing I’m always saying about San Francisco is that if you want to stand out, you need to be weird-maxxing. If you’re fully normal here, you have no alpha.”
Just four other people work at Julius, all engineers. And all of them went to state schools. “It was a coincidence,” Sonwalkar said. “But I was looking for the next Rahuls, who were just undiscovered talent. They’re not people who are just looking to add another company on their resume, especially an AI company. They’re truly into the craft, and they want to build things.”
They haven’t done any sales, Sonwalkar said. “It’s crazy, but people just find it useful, he said (I don’t find it crazy at all). “They put a credit card down and start using it.” The extent of Julius’ marketing appears to be the distribution of embroidered tee shirts, one of which Sonwalkar can be seen wearing regularly.
When he met Nvidia CEO Jensen Huang earlier this year, he was sporting one, just as Huang was sporting his uniform of choice: a black leather jacket.
Sonwalkar recounted their short discussion to me: “I didn’t want to ask him for something small, so I asked him for a speaker slot at the next GTC.” (GPU Technology Conference, Nvidia’s flagship annual event)
“He thought that was crazy,” Sonwalkar said, laughing. His admiration for Nvidia and its formidable leader was palpable. “Nvidia is just a special company, because it shouldn’t have survived. No company survives what they went through in the first decade. But they survived.”
“That’s part of why building companies is really exciting. It’s like the new Wild West,” he told me. “Success is rented, and the rent’s due every day.”
And right now, a million users, three million data visualizations, and 5,000 support tickets later, Julius is paying the rent.
Photography by Anant Sinha
