Tower of Babelbots

One week it’s ChatGPT. The next week it’s Gemini. I can’t keep up. Nor, I’m afraid, do I care. Recently, I set a few of the models to talking and left the room. When I returned, they were stuck in a conversational loop so deadening that I unsubscribed from the lot of them then and there, disgusted. This is what humanity has worked 5,000 years to achieve? A tower of Babelbots? No more “premium” AI for me.

Of course, this puts me in the vanishing minority. Humans and businesses everywhere are AI-ifying at what can only be called an inhuman rate, to the point that even the technology’s creators, the software engineers, are finding themselves out of work. For those who remain, it’s AI or die. The only question is, which model should they use as they auto-obsolesce?

I might not care about the answer—they’re all, pretty much, the exact same—but I do care about the human dramas involved. What must it be like for OpenAI-ers, for instance, as their rivals at Anthropic crush them at coding agents? Not so fun, reports Maxwell Zeff. He’s our newish AI writer—that’s a writer who writes about AI, to be clear, not an AI who writes—and this is his first WIRED feature. I hope to read many more.

It’s funny. When we assigned Max this story a couple months ago, Claude Code was dominant, while OpenAI’s Codex barely registered. But just the other day, a coder friend texted me this: “Things are evolving too quickly. Apparently in the past week or so ppl are preferring Codex 5.3 over Claude Code.” For a brief moment, I was amazed. Then I remembered the tower of Babelbots. What are we even talking about anymore? I’m not sure anybody knows.

 READ THE BIG STORY ➔ 

Likes, Views, and the American Political Machine

In a recent installment of WIRED’s Trendlines video series, politics writer Makena Kelly walks us through the long history of social media influencing and explains how right-wing content creators have come to dominate and drive politics. These creators, Makena says, “use their audience’s loyalty to fuel political narratives, all while profiting off of the outreach.” 

In 2022, Ben Wofford reported on the early days of this exact phenomenon. He homed in on Urban Legend, one of the first influencer marketing firms to be focused on delivering policy messaging directly to followers. Near the end of his feature, Ben revealed that WIRED had located and analyzed 700 posts from the firm’s influencers, posted over the course of five months. They proved that Urban Legend’s campaigns were far more political and incendiary than the firm claimed they were. Today, four years after Ben’s piece, these types of posts remain unregulated. Read his feature and share your thoughts. What will the future of political influencing look like? Send me an email or comment below the story.

In this week’s Big Story, Maxwell reflects on the consequences that could result from shifting the bulk of the programming work—work underlying so many of our essential systems—to an army of corporate-built bots. In the comments, readers had some thoughts about the situation. One is frustrated with the proponents of AI coding, pointing out a callous quote from one of Maxwell’s sources: “Societally, who fucking knows what this means … but I’m pretty optimistic about what’s happening.” Another believes that we aren’t thinking about coding holistically enough. “The public does not realize what goes into creating a computer-generated system,” they said. “Code generation doesn't just happen. A longer period of time goes into systematically outlining a system's planning … AI is a tool, not a planner. There is too much of a human factor involved in planning.” 

Tell us about your favorite WIRED stories and magazine-related memories. Write to samantha_spengler@wired.com, and include “CLASSICS” in the subject line.