I regularly ask media executives what their business looks like in five years. I’m not looking for a prediction, but I do believe it helps to have hypotheses. The Atlantic CEO Nick Thompson and The Guardian CEO Anna Bateson gave me their views at TRB Audience Summit in September.

That led us to create Media 2031, a project where we will look ahead to what a media and information ecosystem transformed operates in five years. My own hypothesis is that, no matter how much my AI delivery agent TJ annoys me with its wild stabs at proactivity, AI will be the front-door to information but also play an increasing role in its creation. I believe publishers will see their ad businesses become smaller. More will use their brands as the front end to different businesses that sell access to unique datasets or IRL gatherings. Text will rarely be the core output as those that thrive will all be multimodal.

Over the next six weeks, we’ll publish a series of conversations about Media 2031, starting today with a level-setting conversion with Axios media correspondent Sara Fischer about how the open web will become the domain of agents, Penske Media's Ryan Pauley and Raptive CSO Paul Bannister.

It all builds to Dec 8 in New York, where we'll both celebrate the end of the year and have an informal town hall-style conversation built around provocations on what Media 2031 will bring

We'd like to hear your provocations. We aren't looking for predictions so much as hypotheses. For instance, the case that websites will mostly be made for agents and humans will rarely visit them. Or even that this time (no really, this time) email will actually be dead.

Send us yours in 100 words or less. If we run yours, you’ll get an invitation to our Media 2031 gathering.

Thanks to our Media 2031 partners: Viafoura, Raptive, DoubleVerify, ProRata and Pathos.

Sara Fischer, media correspondent at Axios, kicked off Media 2031 with a tour of the trends she's tracking. The one that stuck with me is what happens to the open web once AI becomes the front door.

"The open web will be owned by agents,” she said. “The open web is for agents to scrape agent-made pages."

That’s why Sara is tracking closely the licensing deals publishers are negotiating with AI companies. Licensing will be a far larger part of the typical large publisher P&L. At People Inc, for instance, licensing grew 49% last quarter compared to a year earlier.

Other changes Sara is tracking:

Listen to my conversation with Sara

Bloomberg’s Joe Wiesenthal has a great clip where he is equal parts bemuses and amused by how the tech elite keep stepping on rakes when they stray from their bounded world of apps and code and into the messy world of politics. “No intuitions,” he said.

My belief has always been that tech people are pattern matchers and a herd species. That’s why they tend to be original thinkers who happen to land on similar positions. This leads them to project from their bounded areas of code into unbounded, far messier areas like politics. It results in Elon Musk blowing nearly $500 million and moving from blue California to now blue Texas. On the plus side, it has given us tremendous content in the form of All-in struggle sessions by first-principles thinking asset owners.

I’m overthinking it, according to Troy Young. "These are nerds from high school, all of these guys,” he said on the latest episode of People vs Algorithms. This week proved it. Nvidia’s Jensen Huang now wears leather jackets around and signs boobs because he's out as a star, and he was reportedly the target of a Chinese spy. These guys are celebrities now, and celebrities get scrutinized. Their private lives get dug into. You don't get the invites to the White House without the downside.

The bet on Trump was the result of hoary first-principles thinking but basically just opportunism mixed with social climbing. They liked the attention. They liked the invites to the White House. They liked playing the Big Man. It sure beats being the growth guy from Facebook who hustled dubious Spacs.

I don’t really blame people that wanted to build businesses, although there were times when morality went out the window in some of the decision making,” Troy said. “I feel like the terrarium is broken.”

The AI soap opera getting a sex scandal and spy thriller wrapped into one is a great plot twist. Tech hasn’t yet recognized it’s part of the Washington-centered power complex. And now it will get the Washington scandal treatment. If the tech guys don’t like the direction Wired has taken, wait until they are on the receiving end of the press with a wounded administration and a juicy scandal. Expect this to quickly land on the standard: What did he know, when did he know it?

Silicon Valley has long operated as its own universe with its own mores. Some of those mores are, well, out of step with most of the rest of the country. It’s in California, after all. But with great power comes not just great responsibility but great scrutiny – and vicious power moves. Travis Kalanick, who has spent time in the barrel, advises Jensen to get on war footing.

All of this is happening at a fraught moment for both the AI trade and the AI project. It’s best to separate the two. I’ve lived through enough bubbles to know that financial bubbles are the norm, but what matters most is whether the tech is delivering productivity growth. AI trade aside, AI’s massive investments need commensurate productivity growth outside of coding. That was a great trick. Chapeau. But these agents, at least from my experience, are often failing.

The consumer agents often get stuck. I asked Muse to set up time at the Apple Genius Bar, only for Apple to then lock my account because of suspicious activity. I then couldn’t even make the appointment on my own, which would have taken a couple minutes. On the work side, I find Claude Duet a thirsty Odie from Garfield: well-meaning, tongue hanging out with repressed energy, and often creating problems by not understanding how humans think and operate.

The last point is the most troubling one for me. AI has proven best at bounded tasks and worse at fuzzier ones. That’s why AI is so good at coding but still bad at writing. It doesn’t do as well in unbounded areas with lots of shades of gray. Just like the tech guys misjudged the vast and variegated society we have, the AI project will depend on the tech finding a happy middle ground with humans.

Listen to People vs Algorithms

Also, we were joined by Anthony DeRosa to discuss the Media Lab on PvA Field Notes and his experiments with AI. I’m a firm believe that, as frustrating as many tools can be, there’s no replacement for getting your hands dirty with the tools to find what Ehtan Mollick calls the jagged frontier of capabilities. Watch PvA OT with Anthony DeRosa.