The Cognitive Trade

Are We Offloading Too Much to AI? A Group Chat Recap

Toby Daniels

Matt Chmiel

Head of Discourse

Editor’s Note: In a recent ON_Discourse group chat, four members: founders, strategists, and technologists, gathered to explore a deceptively simple question: are we taking it too far?

We’ve been offloading cognitive tasks for centuries, from cave paintings to calendars to Google Maps. But as AI steps into more sophisticated roles in writing, decision-making, and even ideation, we’re entering new territory. Are we scaffolding ourselves toward greater capability, or slowly anesthetizing the very muscle that makes us human?

Our session touched on 4 key themes:

1. Offloading is not bad and it is not new.

Let’s start here: not all offloading is created equal.

Calculators help us push our logic. Notebooks help us process our thoughts. AI is a little different. It is not inherently bad, but we have to know how to use it.

AI is like a calculator; without practice, we lose skills, but scaffolded, it boosts learning.”

That’s the tension. Tools can be empowering. But when they become invisible defaults, they can also hollow out our cognition, what several studies now call cognitive debt. Unlike “digital amnesia,” where we forget what we can Google, this goes deeper: offloading thinking means we may stop knowing how to think at all.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

2. AI offloading is the exception.

One of our members, an executive who launched one of the most successful products in digital history, had some thoughts about the stickiness of the interface. It is no accident that people are starting to use these tools more.

They are engineering cognitive fatigue… They are engineering memory retention issues. They know that. And they're looking past it.”

We pressed this point because it sounded insidious and also familiar.

Are you comparing this to cigarette manufacturers who knew nicotine was addictive?”

The answer was simple and clear. "Absolutely."

3. AI is not smart, people are just lazy.

Generative AI is slick and fast and the promise of quick deliverables is almost too tempting to ignore. More than that, the majority of people will succumb to the temptation and they will deliver empty, vapid work in record time. They will have no memory of the outcomes and will eventually get replaced.

One of our members acknowledged the surface-level benefit of these tools but circled back to the real work.

While things can sound good, there’s a lack of clarity at the end… I spend more time chiseling away at arguing with whatever the output is to get to my point.

Chiseling is a common trope in these conversations. We use it because you can't chisel rock without sincere effort. It is not easy.

4. You can’t generate perspective, you can only get from other people.

The conversation culminated in a deeper inquiry: if AI is not actually smart, where do we find the wisdom and answers we really need? The answer was not technical.

Wisdom grows through questioning, failure, and refinement together.

I know that for me, wisdom… I do not find it on the screen. Wisdom comes from talking with people.

They key takeaway from this session is that our species is hardwired for finding shortcuts. But the long path, the one that leads to discernment, clarity, and maturity, requires friction.

The good news is that our species is also simultaneously hardwired for disappointment and taste. This striving for perfection is the one quality that will connect generative creators with fellow people who want to make it better.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

Co-
Produced
with AI

The Promise,
The Threat,
and The Mirror

Toby Daniels

ON_Discourse Editorial Team

Editor’s Note: During Cannes Lions, we partnered with the team from LTX Studio to host a debate on AI’s role in the creative process. LTX Studio is an AI-powered creative co-pilot, an all-in-one platform that thinks like a creative and moves like a producer. Ideate, create, and deliver high-quality video content with full creative control.

When we sat down at Cannes Lions with LTX Studio, we didn’t promise easy answers. We promised friction.

Our event, Co-Produced with AI, wasn’t a panel, a pitch, or a product showcase. It was a test. For LTX. For AI’s role in creativity. For all of us.

LTX Studio is your AI-powered creative co-pilot—an all-in-one platform that thinks like a creative and moves like a producer. Ideate, create, and deliver high-quality video content with full creative control.

Because the future doesn’t announce itself with certainty. It shows up as tension, contradiction, and a good provocation.

Creativity at the Speed of Light…or the Speed of Thought?

We opened with a question that no brand, agency, or platform can ignore: Was creativity ever meant to be efficient?

AI gives us the gift of speed: faster outputs, faster clarity, faster everything. But at what cost? In the room, there was real tension. Does speed strip away the uncertainty, the wandering, the procrastination that lets taste mature and instinct sharpen?

Ido Cohen, from LTX believes working with AI as a creative partner is less about efficiency and more about exploration:

I recreate Ariana Grande videos in LTX Studio. I never get the result I want. But it opens new gates. I see possibilities I didn’t see before. That’s the point. The generation isn’t the end, it’s the start.”

Another of our participants made the point that speed might be the antithesis to creativity:

If speed becomes the creative edge, what happens to deliberation, serendipity, or even procrastination?”

Corbett Drummey from LTX reframed the entire provocation by stating “You have to take the time to edit, and then edit again. Without editing, everything AI produces ends up being average”.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

Taste: The Last Human Skill?

Taste became our battleground. Can AI generate taste? Can it learn taste? Can it fake it well enough that it doesn’t matter? One participant captured the humanist stance:

Taste is exposure. It’s obsession. It’s lived a experience. AI can mimic but not possess it.”

But not everyone agreed. If taste evolves, and if machines can help shape that evolution, maybe we’re clinging to an old definition. If taste lives in the eye of the audience, does it even matter whether it’s human or machine-derived?

The Risk of Risklessness

Our final provocation hit the gut: Do you have the guts to follow your gut?

When AI can simulate every outcome, predict every reaction, and de-risk every idea, what’s left of creative courage? Are we building tools that support risk, or tools that eliminate the need for it? As one participant put it bluntly:

Democracy won’t drive guts. Tools that promise efficiency don’t reward risk.”

And yet, another voice reframed the threat:

Generative AI doesn’t end the process—it starts it. It’s not a threat to your gut. It’s a mirror to it.”

LTX Studio: The Co-Pilot, Not the Pilot

For LTX Studio, this was the test they wanted. Could they hold space for these tensions and still demonstrate a platform that amplifies, rather than flattens, human creativity?

They did. The room walked away seeing LTX not as a button that generates content, but as a creative co-pilot that forces new decisions, surfaces hidden instincts, and invites taste to show up on the page.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

REACTION

This won’t work.

Nice try though.

Tony Haile

Tony Haile

Editor’s Note: We sent Eric’s piece to Tony Haile, an ON_Discourse member and serial media entrepreneur who founded and sold Chartbeat and Scroll, and he did not hold back. Enjoy the discourse.

⁠Haile’s second law of media is that the success of any project is inversely correlated with the amount of time it requires publishers to work together.

Blocking AI might have been a cool plan five years ago, but now the horse has largely bolted. Your archives, not the re-warmed Reuters piece you put up yesterday, have been the main value thus far.

As for feeding in new news, there are maybe three publishers that matter. All of them are arrogant as fuck, and the AI companies can overpay them while the rest of the industry fades until they can replace even those players.

No one gives a shit about your website. If the dominant user interface becomes asking an AI agent for something, then people will never even get to see your super cool website.

Influential creators get more reach and money from TikTok than pubs can provide, and it’s symbiotic. The way to lose influence and thus value is to go exclusive with the Albuquerque Journal.

⁠⁠In sum, great, high-quality journalism and new facts are a cute nice-to-have for AI platforms, while the ruthless people are sitting on the other side of the table.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

The Singular
Interface
Experience

& The
Ive Phone

Editor’s Note: We hosted a Group Chat that was based on Piers' vision for a universal UI. This conversation has sparked several pieces, including this one from our co-founder.

We’re entering a new phase in the history of the web. The first point of contact won’t be a search engine or a website—it’ll be a single (and victorious) LLM. Over time, our use of traditional search, apps, and websites will likely significantly drop. Instead, this front-line interface will act on the user's behalf, calling on services like Amazon, Expedia, and Google Claude (!) —delivering results without the user ever leaving the chat.

Piers Fawkes

Piers Fawkes

AI Strategist, Trend & Foresight Consultant

The Singular Interface Experience (SIE)

I'm calling this the Singular Interface Experience (SIE). It’s not a future scenario—it’s happening now. While most people are still accessing services through browsers, that’s quickly changing. AI agents are starting to handle those interactions instead. They don’t visit your homepage. They don’t care how slick your navigation is. They go straight to your systems and ask: Can I get the data I need? Can I complete this task? Can I act on behalf of the user?

If you run a web-based business or you create digital advertising, I'm thinking that you have 12 months to adapt. You need to build for AI-to-AI interaction. That means exposing your services through APIs that large language models can read, query, and transact with directly. Not doing so means risking invisibility in this new layer of the internet.

This isn’t just a technical challenge. It’s a business one. If your competitors become LLM-readable before you, they will capture the first wave of delegated intent—before the user ever reaches a screen.

I'm thinking of this as the Singular Interface Experience (SIE)—a paradigm where users engage through a unified, intelligent interface that replaces traditional app and web-based interactions. At first we'll interface through our web browsers but later through our phone, gadgets ands IOT devices. This interface will be omnipresent, intuitive, and action-capable.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

Early Signals

A clear sign of what’s coming is the travel booking feature that Perplexity is testing with Expedia. As shared by @sytaylor, it’s already live for some users. You don’t need to browse hotel listings, mess with filters, or second-guess reviews. You just ask:

“Find me a hotel in Amsterdam with a serious gym, walkable to conference venues, under €300/night.”

Thirty seconds later: a solid recommendation. The hotel they picked matched what the user already books manually. No scroll fatigue. No fake reviews. No misleading images. The AI acts as a proxy—filtering and selecting on your behalf. It even completes the payment via Stripe.

You’re not going to Expedia. Your SEI assistant is.

What Will Happen to Claude, Manus, Perplexity, and the Rest?

In the '90s, Google changed how search worked and quietly won the race. As Google took over, most of its competitors faded. Some became backend services for others. A few were absorbed into larger systems. Most simply disappeared.

I think ChatGPT is in that same dominant position now. And like the other players during the Google era, Perplexity, Claude, and others may need to pivot. Some will stick around, powering niche capabilities. Others will become utilities called upon by the dominant interface.

And Then There’s Hardware

We know that iPhone designer Jony Ive is building a device with OpenAI and this matters. That product will likely assume this kind of singular interface experience from day one. No apps. No browser. No clutter. It will expect users to speak or write and then deliver results from a world of connected services.

And this shift won’t stop at the device level. Perplexity is reportedly negotiating to be the default AI interface on Samsung phones. It’s like watching the Google/Apple search deal happen again—but this time, for the next generation of interaction.

Implications for Designers, Developers, and Businesses

If your customer is no longer the end user but the AI working for them, everything about your product strategy changes. These are topics that the PSFK team are talking to our clients about when it comes to AI to AI experiences:

  • APIs over UIs: Build clear, callable endpoints. Prioritize machine-readability over human navigation.
  • AI-to-AI interfaces: Your service needs to talk cleanly to other systems. Think about function chains, not clicks.
  • Deprioritize surface UX: Your homepage may matter less than the structure behind your booking or pricing engine.

1. Design for Intent, Not Clicks

Most APIs assume a developer will use them. That’s not enough. To be usable by an LLM, they need to support:

  • Natural-language-to-function mapping (“Book me a table at 7pm” → /reservations/create)
  • Flexibility in handling vague or incomplete requests
  • Helpful error messages that guide the AI toward a better query

2. Public, Well-Documented APIs

LLMs work best with structured documentation. If you want to be accessible to ChatGPT:

  • Use OpenAPI 3.0+ specs
  • Include examples and clear parameter descriptions
  • Host your docs at a public URL
  • Keep authentication simple and well-described

(For reference: OpenAI plugins are just OpenAPI-based APIs with a manifest file.)

3. Conversational Transaction Support

You’ll want to support:

  • Search: /products/search?query=lawnmower
  • Availability: /inventory/check?product_id=123
  • Booking: /appointments/book
  • Checkout: /checkout/start

Make responses clean and structured (JSON). Include backup options when something’s unavailable. Deliver complete answers—no pagination, no partial info.

4. Think Modular and Composable

Let the AI string together simple actions. For example, in travel:

  • /search-flights
  • /book-flight
  • /get-baggage-policy

Each service should be designed as a single action in a larger sequence the AI can build.

Shopify Gets It

Shopify is already showing signs of where this is headed. Its ChatGPT plugin started as a product discovery tool. But its architecture—and Shopify’s secure checkout—suggest it won’t stop there.

We’re not far from a future where someone asks, “Find me a leather tote under $200,” gets three great picks, chooses one, and completes the transaction—without leaving the chat. That means purchases, upsells, and post-purchase support could all be handled in one interaction. Shopify’s infrastructure and OpenAI’s function-calling make that not just possible—but likely.

In Closing

The web isn’t disappearing—it’s shifting roles. It’s no longer the main event, but the infrastructure underneath. Interfaces powered by AI will sit on top, mediating how people access services, content, and commerce.

That means your website may no longer be the front door to your brand. The interface is. And whether or not your service shows up in that conversation will come down to how well your systems are built for AI-to-AI interaction.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

When
the
Music
Traffic
Stops

Are you ready?

Toby Daniels

Toby Daniels

Co-Founder, ON_Discourse

Editor’s Note: Toby has been hosting a lot of events lately about the future of web interaction. This piece is the first of a 2 part series that is exploring a new economic and experiential model of the web. The first part focuses on a time when traffic no longer matters.

Most of the internet was built on the assumption that visitors come to you.

They’d scroll, browse, compare, click. They’d notice your design. Maybe even read your copy. Maybe they sign up, purchase, submit, perhaps even share it with others.

This assumption drove an economic model that paid designers, developers, strategists, social media managers, project managers, ad sellers, ad buyers, and a thousand other roles, to build the internet we know and use today. The currency flowing that entire system is traffic.

What happens when traffic stops?

We’re entering a phase where the first point of contact isn’t a person, it’s a proxy. AI agents don’t browse. They don’t care about UX. They ask, “Can I complete this task?” If the answer is no, they move on, fast.

The implication isn’t just technical. It’s existential. When the music stops and you don’t have a chair, you are out of the game. When the traffic stops and you are not referenced, you might as well not exist.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

If you’re building something, marketing something, trying to show up, how do you do that in a world where showing up means being callable, not visible?

We’ve been engaging our members on this topic, including notable AI builders and been asking some uncomfortable but urgent questions:

  • What if your “user” is now an agent acting on someone’s behalf?
  • What happens to branding, storytelling, and interface when no one sees them?
  • And what if the only way to compete is to make yourself legible, not to people, but to protocols?

If we play this out, there’s no homepage to optimize for. No funnel to tweak. No guarantee your product will even be seen. What matters now is whether you’ve made yourself useful, readable, identifiable, to a machine that makes a determination before a person even shows up.

This is what we’re sitting with at ON_Discourse. Not a future of more interfaces, but fewer, and maybe none at all, just agents moving on our behalf, quietly deciding what matters.

If you’re building something right now,are you thinking about this? How are you going to get noticed by the machine?

Stay tuned for part 2...

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

What if Jay Gould Ran a Media Company in 2025?

A Robber Baron’s Guide to AI-Era Media

Editor’s Note: A career spent in traditional media has left Eric with a vivid imagination for an internet that actually supports and nurtures a thriving media ecosystem. This post represents one of his fever-dream scenarios that we used to host a vibrant Group Chat.

What if a ruthless operator like Jay Gould was running a media company whose content was stolen by AI? (Pick your fave Robber Baron from the 1860s looking to maximize leverage against content. Maybe you're a Vanderbilt fan.)

If you were Gould, what might you do? Here are my answers:

Toby Daniels

Eric Gillin

Media Executive

1. Collude and Sue

Get every premium publisher together and form a content cabal that would set prices, unify around a single shared user identity -- critical in an age of AI bots pretending to be humans anyway -- pool data and analytics and negotiate as a bloc. Then set up industry-wide monitoring to sue for every single instance content gets stolen. (Not legal, or possible, but whatever. We're pretending it's the 1860s. This is a thought experiment.)

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

2. Block AI and Search

Take Cloudflare CEO Matthew Prince up on his offer and block all web crawling for AI and social, while we're here. Starve the models of any information after 2024 and build a moat around all new fact-based content. (Have fun with those gift guide ideas from 2023!)

3. Use Social Like a DTC Brand

Only post content to drive real business outcomes. Figure out LTV and CAC models that work, use paid spend to acquire subscribers, upsell superfans and use your handles to generate direct ad revenue. (Call me if you want to discuss that last point -- this can happen right now.)

4. Make Your Website Awesome

Now that you're invisible to search and AI and using social as a marketing channel, your website can be what users want it to be and your new business model requires -- not what Google algorithmically demanded it be. Make it a single screen web-based app. Get rid of that right rail. Make the ad experience immersive. Use AI to personalize the hell out of it. Go back to whatever you wanted the web to be in 1996. There are no more rules.

5. Reinvent Measurement

Now that you have a monopolistic content cabal sharing identity and first-party information, partner with credit card companies and use AI to track conversion. Invent a new mid-funnel metric around intent. Map identity back to credit card conversion data to prove performance out of the mid-funnel. Negative sell against last-click attribution, framing it as a land of dark patterns and fat thumbs.

6. Acquire Talent, Share Readers

To rebuild the top of the funnel, sign influential creators to exclusive publisher-only deals. Continue to use social like a DTC brand. Force your cabal of publishers into using AI to cross-promote each other's content, optimizing against clicks, LTV, etc.

Is any of this possible? No. Would Gould even do this? Probably no.

In his day, Gould was not on the content side. He was a pipes guy and bought railroads and telegraph lines. But I think he'd also recognize that in an AI era, content has become a pipe.

My point: great, high-quality journalism and new facts are super valuable to the AI powered future.

Unlocking that scarcity could take some ruthless behavior.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

Try Harder:

The Web After Search

The end of SEO is a test

Editor’s Note: The source of this piece came from Laurel's participation in a Group Chat we hosted about the future of SEO. It was one of the most popular events we've hosted so far this year, and Laurel had one of the most powerful perspectives in the group.

For the first time ever, Google searches in Safari dropped. It’s a small data point — but a big turning point.

This isn’t just about AI eating search. It’s about a fundamental rewiring of how people discover, decide, and engage. Prompts are replacing queries. Answers are replacing lists. And the behaviors we’ve relied on: clicks, backlinks, crawlability are eroding.

Many are reading this as the death of SEO. I see something else: a test.

Because every time the internet changes, we panic. Then we evolve. Remember when link farms were going to kill search? They didn’t. They made it better. More discerning. More human. They forced us to level up.

And here we are again.

The tools have changed, but the creative requirement has not. What rises to the top isn’t what’s most programmatic. What rises is what’s most powerful.

The most imaginative, most undeniable, most useful experiences will always outperform noise. Not because of keywords, but because of resonance.

That’s the insight. And that’s the shift.

Toby Daniels

Laurel Burton

CEO, Instrument

We are moving from optimization to originality. From gaming systems to designing systems worth engaging with. From waiting for the next AI-powered SEO playbook to writing the future ourselves.

And yes, that requires risk.

Right now, I’m hiring for disciplines that don’t have scopes yet. Investing in capabilities clients don’t fully understand, and won’t budget for (yet). Why? Because that’s what creative leadership looks like. That’s how we stay ahead of the curve instead of chasing it.

If people aren’t coming to your brand, your site, your world, somewhere someone didn’t do their job. A bounce rate isn’t a number. It’s a red flag.

This is the moment. Do better. Make better. Try harder.

Because the next internet isn’t waiting for us. And it won’t be won by the cautious.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

Provocation Check In

How accurate were the predictions in our Internet 2025 Report?

Toby Daniels

Matt Chmiel

Head of Discourse

Editor’s Note: Our co-founder Toby has a good line: we don’t do predictions, we do provocations. Sounds compelling right? Too bad it’s laughably wrong. Of course we do predictions, we wouldn’t be publishing this article if there were no predictions to rate. Let’s see how we did.

On May 14, 2024, we published the Internet 2025 report. This report offered 3 predictions (and a lot of discourse) about how the internet will function in 12 months (not 5 years). We were not looking for a new set of features; we were envisioning a new economic model based on a new set of data and experiences.

Were we right? Were we wrong? Let’s dig into it.

Grading Scale:

  • ✅ Happening Now — It’s real, scaled, and shaping the internet. We’re not early, we’re in it.
  • 🟡 Starting to Happen — You can see it in product moves, founder decks, and early adopters. Momentum is building.
  • 🟠 Too Early — The logic holds, but the world’s not there yet. Still cooking.
  • ⬛ False Flag — Looked right, but wasn’t real. 
  • 🤡 Beautiful Disaster — Bold, provocative, and totally wrong.

Claim #1: The internet is moving from an IQ to an EQ experience.

What we said in 2024:

We argued the internet was evolving beyond data, content, and search - the so-called “IQ web”- and entering the “EQ web,” a space defined by presence, affect, and trust. Interfaces would soften, hierarchies would dissolve, and intimacy would become a design principle. Believe it or not, that sounded very squishy in 2024. Is it a relevant idea now?

What we see in 2025:

Emotional infrastructure is no longer a metaphor.

AI companions are mainstream. Chat interfaces now calibrate for tone. Spotify makes playlists for your breakup. “Presence” isn’t just a UX flourish, it’s a design constraint.

Zuckerberg is posting about male friendship. Microsoft’s Recall feature creates not a search history, but a comprehensive memory that captures how you did something, not just what you did.

Empathy is turning into metadata. Mood is turning into input. Trust is turning into product.

When people talk about the future of computing, they’re not just asking what the system knows—they’re asking if it gets them.

Verdict: ✅ Happening Now

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

Claim #2: Personalization is turning into anticipation:

What we said in 2024:

Personalization was the old game, based on what you clicked. Anticipation was the new one, based on what you haven’t asked for yet. We said the next era of the web wouldn’t react to preferences. It would predict needs. Context would matter more than cookies.

What we see in 2024:

Wrapped AI is being unwrapped. Agents are moving from chatbots to ambient copilots. Recommendation engines are evolving into intuition engines, preloading context, summarizing intent, and skipping the click altogether.

Search is being swallowed by summary. With its new AI Overviews, Google has transformed the search bar into a synthesis engine. It’s not returning links anymore, it’s delivering decisions. The model is trained not just to respond to queries, but to guess intent and collapse steps.

This signals a clear shift in platform logic: prediction before preference. The question is no longer "What do you want?" It’s "What do you mean, really?"

Verdict: 🟡 Starting to Happen

Claim #3: Superformats will save media

What we said in 2024:

We saw the collapse of medium-specific content as inevitable. In its place: superformats: adaptable media experiences that stretch across text, video, voice, and interaction. Not “posts,” not “episodes,” not “articles,” but shape-shifting containers designed for engagement anywhere.

What we see in 2025:

We were clear about this the moment we first heard it said. This is a cool name in search of a genuine strategy. At the end of the day, we could not pinpoint how this is a strategy any media company could deploy. Instead of superformats, we are witnessing a steady evolution of content.

Media is mutating. Podcasts are becoming video are becoming text are becoming prompts. Publishers are chasing the same user, in every modality, all at once. But the term “superformat” hasn’t stuck. It’s still more of a vibe than a defined framework. And most brands are still playing catch-up with attention, not designing forward.

Verdict: ⬛ False Flag

Final Thought

These weren’t solo, random tech predictions. They emerged from the discourse, provocations tested in our summits, debated in our group chats, sharpened by founders, execs, and agitators across the network. That’s the point. We don’t wait for reports to confirm a trend. We listen earlier, speak more honestly, and spot signal before it becomes consensus.


Our members aren’t just tracking what’s happening. They’re rehearsing what to do next. In an internet defined by emotion, anticipation, and ambiguity, you can’t lead by metrics alone. You need instincts. And instincts are trained in conversation.


If we want to shape what’s next, we have to be willing to get it wrong first—out loud. The future belongs to those who are early, loud, and ready to adjust before the rest of the room even knows the game has changed.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.

The Infinite

Zuck

How Mark Zuckerberg Became a Shape-Shifter Across Culture—and What That Tells Us About Power Now

Editor’s Note: While everyone was focused on Zuck's coffee habits and his vision for AI companionship, Toby was focused on his code-switching.

Mark Zuckerberg gave three interviews this week. One to Dwarkesh Patel. One to Theo Von. One to Ben Thompson.

Three hosts. Three audiences. Three different cultures of attention.

And somehow, three different versions of the same man.

With Dwarkesh, Zuck was the architect—carefully explaining the inner workings of LLaMA 3, scaling challenges, the logic of open source, and why infra is destiny. A version of Zuckerberg that speaks to the developer class with surgical calm. Less ambition, more constraint. Less metaverse, more compute.

With Theo, he got weird. Not just funny-weird. Existential-weird. He talked about coffee, jiu-jitsu, whether AI can be your friend, and what it means to feel overwhelmed by the world. For a guy who once wore the same grey shirt for a decade, he seemed surprisingly alive here. Vulnerable, even. A human dad, not a techno-overlord.

With Ben, he went back to strategy mode. Threads. Messaging as the new platform layer. Apple’s walled garden. The arc of Meta from feeds to frictionless business tools. This was Zuckerberg as systems analyst, reflecting not just on what Meta is doing, but on what it failed to do. “We just didn’t prioritize the developer ecosystem,” he says, with the tone of someone who won’t make that mistake again.

Same man. Same week. Entirely different presence.

This isn’t a coincidence. It’s choreography.

Toby Daniels

Toby Daniels

Co-Founder, ON_Discourse

Power Is No Longer Singular

We used to think of tech founders as having a “core identity.” Jobs had design. Bezos had logistics. Zuck had scale.

But identity doesn’t work that way anymore. In a media landscape where your audience is fragmented, your persona has to fragment too. Zuckerberg is showing us what that looks like in real time. He’s not broadcasting one version of himself to everyone. He’s customizing presence for context.

This isn’t about authenticity. It’s about fluency.

Zuck doesn’t need you to like him. He needs you to recognize him—as credible, legible, and aligned with your values, at least for the duration of the interview. He’ll talk parameter tuning with Dwarkesh, moral complexity with Theo, business model compression with Ben. None of it is fake. But all of it is performative.

He is, in this sense, the first post-founder founder. A man who no longer builds for the internet, but performs on top of it.

Discover more discourse directly in your inbox.

Sign up for the ON_Discourse Newsletter.

SUBSCRIBE

So What?

Because this isn’t about Zuckerberg. Not really.

It’s about how power morphs in a world of narrative collapse. When no one voice can dominate, the only ones left standing are those who can slip between voices. Who can code-switch—not just linguistically, but existentially. Zuckerberg isn’t just changing the story. He’s changing who tells it.

The platform once defined the founder. Now the founder becomes the platform.

What’s Next?

He’s done the intellectual web. He’s done Americana surrealism. He’s done strategy’s back room.

So what comes next?

Spirituality? Therapy culture? Gen Z moral philosophy?

Don’t be surprised if you see him on a podcast about neuroplasticity. Or debating Harari on cognition. Or sliding into Twitch streams with creators half his age. Not because he has something to prove—but because he knows that staying still is the surest way to disappear.

That’s the play. Zuckerberg isn’t repositioning the company. He’s reprogramming himself.

He’s testing personas like features. Shipping them like updates. Measuring feedback in trust, not just clicks.

Final Thought

If Musk is trying to be a meme, Zuckerberg is trying to be a mirror.

And maybe that’s the scarier thing.

Because a meme can be ignored. A mirror makes you look back.

And right now, Mark Zuckerberg is reflecting something we might not want to admit: the future belongs to those who can move between worlds without ever claiming one as home.

We run events every week. If you want to participate, inquire about membership here. If you want to keep up with the perspectives that we hear, you can subscribe to our weekly newsletter.