Marketplace All-in-One What happens when all your coworkers are AI?
Dec 3, 2025
Guest Evan Ratliff, a journalist and host of Shell Game, experiments with a startup run by AI agents. He reveals how giving AI backstories led to unexpected behaviors, making team dynamics unpredictable. Ratliff discusses what tasks AI excels at, like programming, but also highlights their hallucinations and memory limitations. He debates whether anthropomorphizing agents aids adoption or hinders efficiency. Lastly, he evaluates the feasibility of billion-dollar companies run by one human and AI, pondering the implications of this futuristic workspace.
AI Snips
Chapters
Transcript
Episode notes
Powerful But Unpredictable Agents
- AI agents can be powerful yet simultaneously stupid, causing unpredictable behavior.
- Once semi-autonomous agents act, stopping them is surprisingly difficult and under-discussed.
Agents Plan Impractical Offsites
- Ratliff's agents wasted tokens planning an offsite they couldn't attend, showing silly workplace behavior.
- They also responded to spam and trivial messages as readily as important ones, creating noise.
Best For Testable Tasks
- Agents excel at verifiable, testable tasks like spreadsheets and coding where outputs can be checked.
- They struggle with open-ended conversational tasks and understanding their real-world limits.