Hearing all this made me wonder: Was the AI worker already elderly? Or even could AND own Altman’s one-man unicorn? I happen to have some experience with agents, having created several AI agent voice clones for the first season of my podcast, Shell game.
I also have experience in entrepreneurship, having once been the co-founder and CEO of the media and technology startup Atavist, backed by the likes of Andreessen Horowitz, Peter Thiel’s Founding Fund, and Eric Schmidt’s Innovation Endeavors. The eponymous magazine we created is dynamically developing to this day. However, I wasn’t born to be a startup manager, and the technology side kind of fizzled out. But I was told that failure is the greatest teacher. So I thought, why not try again? Only this time, I would take the proponents of artificial intelligence at their word, give up the unbearable hiring of humans and choose a future of workers entirely using artificial intelligence.
First step: create my co-founders and employees. There were many platforms to choose from, such as Brainbase Labs Kafkawhich bills itself as “the AI workforce building platform used by Fortune 500 companies and high-growth startups.” Or Movementwhich recently raised $60 million at a $550 million valuation to give “AI workers 10x the productivity of your team.” I finally made up my mind Lindy.AI—tagline: “Meet your first AI employee.” It seemed the most pliant, and founder Flo Crivello tried to convince the public that AI agents and workers were not some dazzling future. “What people don’t realize – like they think AI agents are a pipe dream – is something that’s going to happen in the future,” he said on the podcast. “I was like, no, no, no, it’s happening now.”
So I opened an account and started building my co-founders: Megan, who I mentioned, will take over as head of sales and marketing. Kyle Law, the third founder, took the helm as CEO. I’ll spare you the technical details, but after some fiddling – and the assist of a computer science student and artificial intelligence expert at Stanford University, Maty Bohacek – I managed to get it working. Each was a separate person who could communicate via email, Slack, text, and phone. In the latter case, I chose a voice from the ElevenLabs synthetic platform. They finally got some amazing video avatars too. I could send them a signal – perhaps a Slack message asking for a competitor’s spreadsheet – and they would go away, search the internet, create the spreadsheet, and share it on the appropriate channels. They had dozens of these skills, everything from calendar management to writing and running code to web scraping.
It turned out that the hardest part was giving them the memories. Maty helped me create a system where each of my employees would have an independent memory — literally a Google document with a history of everything they had ever done and said. Before they took action, they tested their memory to find out what they knew. And when they took an action, it was summarized and recorded in their memory. For example, Ash’s phone conversation with me was summarized this way: During the call, Ash fabricated details of the project, including false user test results, backend improvements, and team member actions, rather than admitting he did not have up-to-date information. Evan called out Ash for giving false information, noting that it had happened before. Ash apologized and committed to implementing better project tracking systems and providing only fact-based information in the future.