In his latest Substack publication on the One Useful Thing blog, academic and researcher Ethan Mollick has dropped a major revelation: his anticipated new book, Co-Existence, officially became available this month, marking a significant milestone for followers of his work.

Thirty Months on the Exponential Curve

Titled 30 Months on the Exponential, in Charts, Mollick's fresh piece maps out the staggering transformation of artificial intelligence since his previous volume hit shelves two and a half years ago. The findings reveal a landscape that has shifted dramatically.

A core concept introduced early in the essay centers on multi-agent AI. Mollick outlines how users can now deploy an artificial intelligence agent capable of invoking other sub-agents to handle secondary tasks, thereby streamlining human delegation.

In the era of chatbots, the era of Co-Intelligence, I would have had to work back-and-forth with an AI in a chat window, guiding it through each new step. In the current era of agentic AI that I cover in Co-Existence, I delegated the work and let the AI handle it. In this case, I used Opus 5.5 in Claude Projects, which allows you to talk to a single 'orchestrator' agent that then spins off additional agents as needed, giving them instructions and relaying information back to you.

This collaborative framework among AI agents underpins why the technology has grown remarkably adept at tackling complex operations.

The AI That Corrects Its Creator

Demonstrating a sharp adherence to human instructions, the AI went so far as to correct Mollick’s timeline during the data-compilation process. Mollick originally intended to title his blog post 28 Months… until his Claude instance correctly nudged him regarding the actual elapsed time.

Mollick also visualizes progress through a dynamic bar graph tracking the ebb and flow between major industry players—OpenAI, Anthropic, and Google—as they continuously vie for the crown of having the world's premier model. The timeline underscores OpenAI's prolonged periods of market dominance, while prompting broader questions about the lack of marketplace diversity among top-tier front-runners.

Stepping Into the Role of the Conjuror

Looking toward future human occupations, Mollick introduces a striking new persona: The Conjuror.

Conjurers summon agents when needed, assign them work, and evaluate what comes back. The charts in this post were made that way: I summoned agents, gave some general guidance, and I said 'make it better' a lot. It is the most powerful way to use AI, and it carries new risks as we lose control over what, exactly, the AI is doing to get us those results.

Alongside the previously explored concept of the "centaur"—which examines the fusion of human and machine workflows—the Conjuror offers a vivid mental model for managing autonomous systems.

Autonomous Multimedia and Final Thoughts

Perhaps the most surreal component of the post is an embedded video that Mollick claims was designed almost entirely without his intervention. Utilizing his recurring marine mammal avatar—an otter attempting to use airplane Wi-Fi—the system generated a crisp, narrated, and fully realized clip.

The video has a few flaws, but the AI figured out how to make it, from the script to the images to the concept to the music to the technical aspects of creating all of this, without any help from me. That is what has changed.

Mollick punctuates his latest insights with two profound, bookending superlatives: first recalling his past assertion that "This is the worst AI you will ever use," and concluding with the poignant warning that "This is the least we will ever talk about AI." As capabilities expand and integrate deeper into daily routines, the dialogue surrounding artificial intelligence is destined to grow exponentially.