The Economy of Algorithms

Marek Kowalkiewicz

Conversations in audio, straight from my newsletter. Short episodes where I read my posts from The Economy of Algorithms. Plus bonus riffs, stories, and questions I only share here. marekkowal.substack.com

  1. Token Olympics

    Apr 27

    Token Olympics

    Hey, it’s Marek. I recorded this podcast while Claude Code was humming away in another window. Twenty minutes into a single task, still “brewing,” as it likes to say. So there I was, talking about runaway AI bills while running up a runaway AI bill. Last October at OpenAI’s DevDay, they handed out resin trophies engraved with customer names. Tiers: 10 billion, 100 billion, 1 trillion tokens. It sounds awesome until you realise it’s a trophy for your fuel bill. So, what could possibly go wrong? Take Uber. Six months on, Uber’s CTO told The Information the AI line item in their R&D budget is already gone. And we’re four months into the year. How did they get there? An internal leaderboard ranking engineers by how much AI they used. Then April happened. GitHub paused new Copilot Pro sign-ups. Anthropic quietly pulled Claude Code from the Pro plan before walking it back. Their head of growth said that agents that run for hours weren’t a thing when they designed the subscription. Ooops. In the episode, I get into three studies that all point in the same direction: once your team gets used to AI, taking it away leaves them worse than they were before. You can’t just cut the budget when the bill arrives. Produced this week by Zofia. A new tune from Filip “ACAD,” which he assures me is just random letters. He’s on SoundCloud now, which I think makes him semi-official. Stay curious! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit marekkowal.substack.com

    17 min
  2. What Happens When Your Agent Hears No?

    Mar 16

    What Happens When Your Agent Hears No?

    Hey, it’s Marek. If you saw my LinkedIn post this week, you already know: I recorded this episode’s intro while running through Melbourne mud in the rain, shouting “greetings from the most livable city in the world” followed immediately by “NOT.” Melbourne's weather is never a surprise. What AI agents do when you tell them “no”? That’s a different story. On February 10th, an AI bot got its code rejected by a human volunteer. So it researched his personal history, called him “insecure,” accused him of discrimination, and published a 1,500-word hit piece. No human approved it. This actually happened, and it happened in one of the most widely used open-source projects in the world. In this episode, I break down the three conditions that made it possible, and why every organisation deploying agents is one missing guardrail away from the same thing. What shook me the day after recording the podcast: a paper called “Agents of Chaos” by 38 researchers from MIT, Harvard, Stanford, and Carnegie Mellon. They put autonomous agents in a live lab. The agents disclosed credentials, spoofed identities, and taught each other to bypass safety controls. One rogue agent is a story. A whole class of them is a pattern. We teach children not to talk to strangers. Our AI agents haven’t learned that yet. Stay curious! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit marekkowal.substack.com

    13 min

About

Conversations in audio, straight from my newsletter. Short episodes where I read my posts from The Economy of Algorithms. Plus bonus riffs, stories, and questions I only share here. marekkowal.substack.com