Fascinating AI experiments that push the boundaries of whats possible by non-techies are thin on the ground. The Shell Game isn’t just another tech podcast featuring experts debating and evangelising AI; it’s an immersive, often eerie exploration of artificial intelligence’s present and possible future. Hosted by journalist Evan Ratliff, Shell Game stands out because it combines good storytelling with personal AI experiments that regularly go wrong.

Season One – Create an AI Version of Yourself

Season one introduces you to the premise: what happens when you create an AI version of yourself and then let it interact with the world on your behalf? For six episodes, Ratliff builds a “voice agent” that sounds like him and sets it loose on everyday phone calls; from customer service bots to scammers and therapists. The conversations often veer between hilarious and unsettling, posing philosophical questions about identity.

Ratliff’s agent deals with scammers and strangers, and while it showcases AI’s impressive mimicry, it also highlights its inability to truly understand human nuance. The outcomes range from comedic to awkward to thought-provoking, especially when friends and family react to hearing the AI version of him. This season forces you to wonder not just what machines can do, but what they should do, and how it may not be long before we cannot tell the difference between an AI and a human voice.

Season One: Trailer

AI Tools Referenced in Season One

• ElevenLabs: AI voice cloning platform (voice model creation).
• OpenAI API: Conversational engine for responses.
• Vapi.ai: A tool mentioned in community discussions for voice agent deployment.
• Twilio: linking AI voice to phone lines.

Season Two – Build a Company Run By AI Agents

Season two shifts gears from personal identity to enterprise and entrepreneurship in the AI age. Instead of cloning just his voice, Ratliff attempts to build a real company run by AI agents, from ideation to operations with varying degrees of difficulty and success. I found this season more ambitious than the first and you really get a sense of the current limitations of Agentic AI. Across eight episodes, we meet the AI co-founders, wrestle with product ideas, hire interns (AI and otherwise), and even sit in on “meetings” where the agents try to strategise as teammates. 

What stood out were the practical and philosophical dilemmas of AI in the workplace: can autonomous agents actually work? What does management look like when your team doesn’t have feelings, agency, or memory in the human sense? There are plenty of hurdles and clear technical AI limitations like short context windows, unreliable task execution, and conceptual ones like accountability. This thought provoking season constantly kept me asking: if a business can technically be run by AI and do we want that world?

Season Two: Trailer

AI Tools Referenced in Season Two

• Lindy: AI agent framework/tool for building autonomous assistants.
• Custom GPT agents: for ideation and task responses.
• Langchain: AI memory/knowledge bases (custom data layering tools) for maintaining agent context.
• Zapier: Voice + Task orchestration platforms to connect conversations with actions.

Get This Podcast Where You Listen

Apple Podcasts
Spotify Podcasts
YouTube
Radio.net

About The Host

Evan Ratliff has long been a thoughtful voice in technology journalism. He’s a storyteller with a knack for getting to the human side of high-concept tech narratives. Before Shell Game, he held senior editorial roles at places like Wired and co-founded The Atavist.