Party of one

lone founder

Artificial intelligence now turns the lone founder into a credible rival to the corporation, and the steady salaried job into a riskier bet.

In February 2025 Maor Shlomo, a programmer in Tel Aviv, put online a side project called Base44, an app that builds working software from instructions typed in plain English. Within a month it had earned close to 1.5 million US dollars. Word spread on LinkedIn rather than through any advertising, and the user count passed 250,000 inside three weeks. Shlomo ran the lot himself, waking every few hours to check the servers were still up. The deal to sell to Wix, an Israeli website-building firm, was struck in June for 80 million US dollars in cash. He had hired his first employee only weeks earlier.

Sam Altman, the boss of OpenAI, has mused about this for a couple of years. He told Alexis Ohanian, a co-founder of Reddit, that his group chat of fellow tech bosses was running a betting pool on the first one-person, billion-dollar company. Nobody has collected. Shlomo had a small team by the end, and 80 million US dollars sits well short of a billion. The nearest thing so far is Midjourney, an image service that David Holz has run since 2022 with no venture money and fewer than 200 staff; he told The Information in March 2026 that revenue had passed 200 million US dollars. It sells subscriptions and has turned a profit since its first weeks. The method even has a name, coined by Andrej Karpathy, late of OpenAI and Tesla, in early 2025: vibe coding.

Vasant Dhar, who studies trust in artificial intelligence at New York University, has doubted that a true solo unicorn will arrive soon. Founders, he notes, are loath to hand their riskiest work, such as reading a contract with a major client, to a machine that may slip. Shlomo still hired.

While Shlomo was building, Marc Benioff was cutting. The boss of Salesforce, a business-software giant, told a podcast in September 2025 that he had taken his customer-support staff from 9,000 to about 5,000, because the firm’s agents now fielded half of all conversations. “I need less heads,” he said. Salesforce took on no new engineers in its 2026 financial year, and in February 2026 it cut almost 1,000 further posts in marketing and product. Klarna, a Swedish payments firm, had pointed the way a year earlier, when Sebastian Siemiatkowski, its chief, said an AI assistant was doing the work of 700 agents. Amazon shed 14,000 corporate jobs last autumn, though Andy Jassy, its chief executive, put the reasons down to culture rather than machines.

Erik Brynjolfsson, a Stanford economist, put numbers to the drift in August 2025. With two colleagues, Bharat Chandar and Ruyu Chen, he combed payroll records covering millions of American workers. Entry-level employment in the most AI-exposed jobs had fallen 13 per cent since late 2022, even as older staff in the same roles held steady or gained ground. In software engineering and customer support, the two fields the study singled out, the drop for the young was closer to a fifth. Brynjolfsson called it the fastest and broadest change he had seen in the workplace, short of the rush to remote work.

Hands off

Shlomo has said that for three months he wrote no front-end code at all, because the AI did it for him. His work was to choose what to build and keep the project moving in one direction. Benioff describes the same shift from the other side: he now runs a workforce of people and software agents together, with an agentic sales team ringing back enquiries that his staff, he says, had gone years without answering. Holz keeps Midjourney small on purpose, licensing its house style to Meta last August rather than selling up or swelling the payroll.

In March 2026 Wix’s annual report disclosed the next instalment. Shlomo, who had earlier walked away from the well-funded start-up he founded, was in line for a further 90 million US dollars should Base44 hit its targets.


Photo: Dreamstime.

Privacy Preference Center

Strictly Necessary

Cookies that are necessary for the site to function properly.

gdpr, wordpress_[hash], wordpress_logged_in_[hash], wp-settings-{time}-[UID], PHPSESSID, wordpress_sec_[hash], wordpress_test_cookie, wp-settings-1125, wp-settings-time-1125, cookie_notice_accepted

Comment Cookies

Cookies that are saved when commenting.

comment, comment_author_{HASH}, comment_author_email_{HASH}, comment_author_url_{HASH}

Analyze website

Cookies used to analyze website.

__hssc, __hssrc, __hstc, hubspotutk

Targeting/Advertising

Cookies for provide site rankings, and the data collected by them is also used for audience segmentation and targeted advertising.

__qca

Google Universal Analytics

This cookie name is asssociated with Google Universal Analytics.

_ga, _gid

Functionality

This cookies contain an updated page counter.

__atuvc, __atuvs