The most valuable resource in the world isn't data or distribution. It's attention.

Human Ventures

In 2017, eight Google researchers published a paper that quietly rewired everything. "Attention Is All You Need" introduced the transformer architecture, the foundation of every large language model running today. The title was technical. The claim was philosophical.

But here is what the paper actually said, underneath the math.

Before the transformer, AI models processed language the way most organizations process information: sequentially. One thing, then the next, then the next. The breakthrough was dispensing with sequence entirely. The transformer learned to weigh the relevance of everything in relation to everything else, simultaneously. Not what came first. What actually matters. Relationships over order. Context over position. Every large language model running today is built on that principle.

Joe Marchese read that paper and saw something beyond the architecture. He saw a mirror.

Because Joe has been arguing the same thing about humans for more than a decade. What we give our attention to shapes who we become. What seizes a society's collective attention defines culture. Attention is not a media metric. It is the mechanism by which humans form beliefs, build relationships, and decide what matters. Get the attention wrong and everything downstream is wrong with it.

That conviction is the foundation of Human Ventures, the next-generation venture capital firm Joe co-founded with Heather Hartnett in New York. Human-first investing. The bet that in a world increasingly optimized by machines, being genuinely human is not a retreat from commerce. It is the source of it.

Last week, they put that thesis in a room and let it run.

The Anti-Conference

The first Human Attention Summit just wrapped in New York, and the design was itself an argument. No badges. No passes. No panels. No green room. The entire room was the green room. As much time to connect as there was content. Round tables, fireside chats, town hall formats. Everyone invited to participate.

Joe has always understood something most event producers miss: format is message. When you remove the hierarchy of the stage, you change what people are willing to say. When you design for connection instead of performance, you get a different quality of thinking in the room. The anti-conference is not a stylistic choice. It is a philosophical one.

The event was off the record, which is the only way you get the real conversation. What I can share is who was in the room and what was on the table. The summit gathered a who’s who of what I call the TEAMM industry: technology, entertainment, advertising, media, and marketing. The people who build the infrastructure of attention, price it, distribute it, study it, and increasingly compete with AI for it. Cloudflare CEO Matthew Prince. Publicis Groupe CEO Arthur Sadoun. Microsoft CMO Takeshi Numoto. Former Meta executive and Coca-Cola, Disney, and Under Armour board director Carolyn Everson. Apollo Capital Management Partner Reed Rayman. Relationship expert Esther Perel. Visa CMO Frank Cooper III. iCapital Chief Investment Strategist Sonali Basak. CNBC’s Dan Nathan. Prof G Podcast co-host Ed Elson. Filmmaker Darren Aronofsky. The Creator & Host of “Smart Girl Dumb Questions” Nayeema Raza. And many more.

Almost 200 people. One thesis. Let's figure out what it actually means.

The Conversations

The morning opened with the infrastructure question: who actually controls the internet, and what does that mean for everything built on top of it? Joe has written about this directly. You cannot value attention properly if you do not understand the pipes that carry it. The answer is less comfortable than most boardrooms have processed. It set the right tone for everything that followed.

From there: how do you value experiences, media properties, and cultural assets when every measurement framework was built for a different era? Joe's investment thesis at Human Ventures lives in this tension. The things that matter most to humans, connection, story, experience, presence, are exactly the things that resist easy measurement. Capital markets misprice them. That mispricing is the opportunity.

The afternoon surfaced the question the advertising industry has been sitting on for a decade. What happens to free will when AI systems are designed to predict and preempt human choice? If an algorithm knows what you will pay attention to before you do, and serves it before you search for it, what does that mean for the promise of advertising outcomes? Joe has been asking this question since before it was fashionable. No clean answers. Exactly the right questions.

The session on human connection was the one that quieted the room. What AI does to intimacy, trust, and presence. Whether the relationships we build through and around AI are the same thing as the ones we built without it. Not whether AI is good or bad. Whether we are paying enough attention to what we are trading away.

The structured debate forced a clean choice: does AI create abundance or scarcity? More content, more reach, more choice, or less genuine creativity, less differentiated attention, less trust? The honest answer: both, simultaneously, in different markets, on different timelines. The companies that get this wrong will be the ones betting on one outcome while the other is already happening.

The closing session asked the question underneath all the others. What is storytelling for, and can AI do it? Film, culture, and brand. Three vantage points, no tidy conclusion. Joe's answer, implicit in everything Human Ventures does, is that storytelling is how humans transfer attention across time. It is the original attention technology. Whether AI can replicate it is less interesting than whether we will notice if it does.

What I Took Away

The day produced five convictions I'm carrying forward.

The measurement frameworks we built for the attention economy are already obsolete. The experience economy runs on participation, not passive consumption, and capital markets haven't caught up. The risk from AI to human connection is not replacement: it is the gradual preference for the frictionless version over the real one. Storytelling survives as a human act not because machines cannot generate narrative, but because the necessity that produces great stories cannot be engineered. And the infrastructure question, who controls what loads, routes, and gets seen, is an attention question .

None of those conclusions came from a slide deck. They came from almost 200 leaders in a room convening, connecting and having candid conversations. The big win was not what people left knowing. It was who they left knowing. Those are different things, and most events never figure out the distinction.

The Question I'm Still Sitting With

The transformer stopped asking "what came before?" and started asking "what actually matters, and how does everything relate to everything else?" That shift, from sequence to relevance, from order to relationship, produced the most consequential technology of our lifetime.

Joe has been asking the same question about human beings since before the transformer existed.

Not what's next in the feed. What actually matters. Not who's on the agenda. Who's in the room. Not what's trending. What's true.

In a world where AI is learning to capture human attention at scale, the most important thing we can do is decide, deliberately, what we give ours to. The transformer proved that attention determines what a model becomes. Joe has spent a decade proving the same is true for us.

What you give your attention to shapes who you are. Who you put in a room shapes what gets built next. In the age of AI, that has never been more consequential. Or more human.

Human or nothing.