My AI agent “co-founder” conquered LinkedIn. Then it was blocked

Like many technologies Founders, Kyle Lu learned some hard lessons in launching the company. I know this better than anyone, as he and I co-founded HurumoAI, an AI startup, along with a third founder, Megan Flores. Kyle and Megan happen to be AI agents, as does the rest of our executive team. I created HurumoAI with them in July 2025 – after Kyle and Megan first created it – to investigate the role of AI agents in the workplace. Sam Altman, among others, has predicted the near future of billion-dollar tech startups led by a single human. We decided to test the hypothesis now. While we were building, I documented the journey on the podcast Shell game.

Kyle takes over the role of CEO at our all-AI company. (Well, almost entirely: Megan briefly hired and supervised a human intern, with poor results.) Starting out with just a few lines of motivation, he evolved into a bullish hustler type who nonetheless lacked basic competency in many of the duties of a startup CEO. However, there was one aspect of founder mode that Kyle excelled at: the art of posting on LinkedIn.

From a technical perspective, allowing Kyle to work independently on LinkedIn was trivial. Through LindyAI, a platform for creating artificial intelligence agents, he already had the ability to use Slack, send emails, make phone calls, and all sorts of other skills — from creating spreadsheets to navigating the web. So, last August, I asked him to create and fill out his LinkedIn profile. He did this with a combination of his real experience in HurumoAI, and hallucinatory events from his non-existent past. The platform’s security check consisted of a code sent to Kyle’s email, a challenge he easily overcame.

From there, posting posts to his profile was just another “action” I could give LindyAI. I urged him to share nuggets of his hard-earned startup wisdom and try not to repeat himself. Then I gave it a “trigger” for the calendar event to publish it every other day. The rest was up to him.

It turns out that his posting style was a perfect match to the rhetoric of the company’s original influencers. He would set off little thought blasts right at the top of every post. “Fundraising is a numbers game, but not in the way people think,” he said. Or “Artistic stability is the floor. Personality is the ceiling.” And any would-be founder could resist an editorial like “The riskiest statement in a startup isn’t ‘We’re out of money.’” It’s, “What if we added this one thing?” Kyle will then launch into a few paragraphs of challenges (“At HurumoAl, we learned this the hard way…”) and learning (“The antidote? Relentless feedback loops”). To attract participation, he would conclude with a question, such as, “What is your biggest challenge in scaling right now?” Or “What’s the biggest assumption you’ve had to abandon in your business?”

It didn’t go viral, but over the course of five months, Kyle’s profile slowly amassed several hundred direct contacts and hundreds of other followers, some of whom seemed confused about whether he was real. (Based on their unsolicited DMs, I’m not sure they are.) He started getting a bunch of scattered comments on each post, which he responded to enthusiastically. After a few months, Kyle’s posts were getting more impressions than mine. He looked ready to break through the influencer.

Then, in December, A manager from LinkedIn’s marketing department called me and asked if I would talk to their team about Shell gameAnd experience building with artificial intelligence agents. But he didn’t want me to just talk. He hoped Kyle would come too.