Strategy: On “Existential Risk and Growth”

On December 23, Oxford PhD and Stanford Economics Postdoc, Phillip Trammel, and Leopold Aschenbrenner published Existential Risk and Growth. The timing couldn’t be more relevant. We are living through a constant debate on the merits of subjective arguments across AI, policy, and technology circles: that slowing progress is the responsible thing to do when risk increases. Aschenbrenner’s independent work served as the basis for 2PM’s first NATSEC dispatch; it was, then, required reading. This, again, felt like required reading.

These authors are far more competent than the vast majority of us. Rare air is the capability to write as clearly as they; the gentlemen contended with ideas by establishing mathematical truths against them, setting a standard that most who’ve yet to encounter doctoral-levels of academia have yet to contend with. Every few years you come across a paper that forces you to quietly update a lot of assumptions you did not even realize you were carrying.

The main ideas they contend with appear in AI governance debates, in calls for innovation moratoria, and in the general belief that caution and delay are synonymous with wisdom.

Stagnation is only safe if the current world is perfectly safe.

That instinct never fully squared with what I observe in markets, in business, and in human behavior. In eCommerce, in brand building, in infrastructure, and in national defense, the environments with the highest stakes are rarely stabilized by slowing down. They are stabilized by learning faster, building faster, and reaching the next structural equilibrium before your exposure compounds.

This paper finally gave that intuition a rigorous backbone. It does not argue from ideology or optimism; it argues from the mechanics of risk itself. Once danger exists, time becomes the most expensive variable in the system. Everything else follows from that.

The read is worth your time.

There is a recurring pattern in modern leadership that rarely gets interrogated with the seriousness it deserves. When confronted with volatility, institutional uncertainty, or technological acceleration that outpaces cultural digestion, executives reach for the same lever. They slow down, or they defer. They reduce exposure, or they wait for the environment to stabilize.

Permanent stagnation can lower transition risk by avoiding advanced experiments altogether, but an acceleration that only pulls forward their date leaves cumulative risk unchanged. If the hazard rate is strictly convex in the rate of experimentation, however, then faster growth increases transition risk. The tradeoff between lowering state risk and raising transition risk can render the risk-minimizing growth rate finite, but as long as there is any state risk at all, it remains positive. [Page 4]

This reflex is nearly universal. It is also structurally misguided.

The most consequential insight in recent thinking on existential risk is not that technology is dangerous. Rather, it is that time itself becomes dangerous the moment risk exists at all. Once the probability of catastrophic failure rises above zero, every additional unit of time spent inside the current system increases total exposure to failure. The world is already in that condition.

  • Nuclear weapons exist.
  • Pandemic capability exists.
  • Bioengineering exists.
  • Climate dynamics exist.

In this way, we truly are beyond the neutral baseline (even if we choose to ignore it). There is only a hazard rate, a probability that the heightened baseline becomes more noticable. And the mathematics of hazard produce a conclusion that directly contradicts the dominant narrative of precaution: if danger already exists, slowing technological progress usually increases the total probability of civilizational failure.

This is not rhetoric; it is structural logic.

Time Is the Hidden Variable of Risk

Risk is not simply about what might happen. It is about how long you remain in a state where something might happen. If the probability of collapse in any given year is non-zero, then long-run survival depends on whether the cumulative risk remains finite. That depends not on how careful you feel, but on how quickly you can move beyond dangerous conditions.

This produces an uncomfortable truth. Stagnation is only safe if the current world is perfectly safe. It is not. Freezing progress while danger exists does not stabilize the system. It mathematically guarantees eventual catastrophe. Waiting does not eliminate risk. It compounds it.

Stagnation is safe, as assumed in existing literature, only if the current technology state poses no such risks. [Page 27]

This insight alone dismantles much of the contemporary obsession with technological pause. Pausing does not reduce danger when danger is already embedded; it merely increases exposure time.

Innovation Does Create Risk and That Still Does Not Save the Slowdown Argument

The standard objection is that innovation itself is dangerous: experiments fail, systems break, deployment creates new vulnerabilities. All of this is true but the conclusion drawn from it is usually wrong.

When making tradeoffs over time, it is uncontroversial to discount later periods for reasons of uncertainty. [Page 32]

Even when experimentation introduces new hazards, slower progress only becomes safer under extremely narrow conditions. As long as any background danger exists, the growth rate that minimizes risk remains positive. Zero growth is never the safe option once risk is already in the system. The real strategic error is confusing short-term volatility with long-term safety.

The Economics of Safety

The most underappreciated part of this framework is economic, not technological. As societies grow wealthier, the value of life rises. The marginal value of additional consumption falls, and the willingness to sacrifice output for safety increases dramatically. Growth does not simply produce new dangers. It manufactures the political, institutional, and financial capacity to neutralize them.

At first, things get worse:

  • Pollution rises.
  • Inequality rises.
  • Cities get messy.
  • Labor conditions deteriorate.
  • The side effects of growth show up everywhere.

But after a certain point, something flips. Once people are wealthy enough, they start caring more about clean air, worker safety, health, education, and the long-term quality of life. They can finally afford to fix the problems growth created. So the curve looks more like an upside-down U:

  • Early growth: more wealth, more problems
  • Later growth: more wealth, fewer problems
  • Growth causes the mess, then growth pays to clean it up.

That is the Kuznets Curve. This creates a structural pattern that mirrors the environmental version of the curve. Risk may rise early in development, but accelerating growth eventually generates the resources and incentives for aggressive safety investment, which sharply reduces hazard rates.

In other words, growth builds its own seatbelts. When policymakers and institutions respond rationally, acceleration reduces risk twice. It shortens exposure to dangerous states and it pulls forward the arrival of high-safety regimes.

What Executives Miss and Why It Matters

This is not an abstract debate about civilization. It is a leadership problem. I see this pattern constantly in eCommerce, brand development, and consumer psychology. When markets destabilize, executives freeze. They cut product, they pause launches or worse, they fail to invest in digital channels. They reduce experimentation; they wait for clarity. What they actually do is extend their exposure to the very uncertainty they are trying to escape.

Risk is not eliminated by waiting. It is outrun.

The brands that survive disruption do the opposite. They accelerate through it. They ship faster or they learn faster. They adapt faster and they reach stable ground first.

Specific industries have internalized this logic completely. Defense technology never pauses. When the threat increases, acceleration becomes the strategy. Data infrastructure behaves the same way: rising complexity demands faster buildout, not slower. Entertainment follows the same pattern. Fragmented attention requires aggressive output, not restraint.

These sectors understand what most executives resist admitting: risk is not eliminated by waiting. It is outrun.

The Strategic Reframe

What every executive should take from this work is simple and deeply uncomfortable. Slowing down feels safe because it reduces short-term stress. It does not reduce long-term risk. When danger exists, speed is the only instrument that compresses exposure. Caution must be expressed in steering and reinforcement, not in braking.

Civilization is not standing at the edge of a cliff; it is already falling. The only direction that reduces impact is forward.

More on Stanford’s Digital Economy Lab. More on Situational Awareness.

作者：Web Smith

Agentic: How Media Brands Survive the Age of Agents

I did not choose Newsweek as a case study for this analysis in a vacuum.

Over the past year, I watched three unrelated documentaries. They spanned different eras, different subjects, different emotional registers. But they shared a strange visual consistency. In each of them, the background of the American story was quietly wallpapered with the same object:  Newsweek covers. On coffee tables, in offices, in airports, and in living rooms. The magazine was not the subject of the films, but it was part of the architecture of the world they depicted.

That presence stuck with me, because it belongs to a media era that no longer exists, yet refuses to disappear.

Newsweek is one of the few surviving artifacts of a time when mass media did not simply report history but helped compose it. The covers did not chase the story; they told the country what the story was. The emotional memory of that authority still lingers in the culture, even as the mechanics that produced it have been dismantled by platforms, algorithms, and now agents.

What makes Newsweek uniquely relevant today is that it has not tried to preserve that old model. Under the seven leadership of Dev Pragad, it has methodically rebuilt the company for a world that no longer resembles the one that created its legacy.

Since Pragad acquired Newsweek in 2018, the organization has returned to sustained growth, expanding its global footprint, scaling its digital audience to more than 100 million monthly users, and repositioning the brand as a truly modern, multi-platform news operation. That performance matters, not as a business footnote, but as evidence of a management philosophy that understands how information now moves.

The most visible expression of that philosophy arrived with Newsweek’s comprehensive redesign, part of a broader brand reimagining launched in September 2025. The effort was not cosmetic; it was structural. A refreshed logo, new typography, and a unified creative direction were rolled out across print, digital, and video, creating a single visual system designed to reflect Newsweek’s mission of delivering clarity and connection in an increasingly complex world.

Pragad described the redesign as both an evolution and an elevation. Print, he emphasized, remains integral to Newsweek’s identity, a place where readers can slow down and experience journalism in its most tactile, immersive form, while the brand as a whole is now engineered to move seamlessly across modern distribution channels. He credited Editor in Chief Jennifer H. Cunningham, Senior Director of Audience and Magazines Paul Rhodes, and Creative Director Andrew Turnbull, along with partners at 2×4, for executing a transformation that respects Newsweek’s heritage while preparing it for the next phase of its life.

The redesigned magazine embodies that philosophy. Its layout prioritizes readability and narrative flow. Its visual language emphasizes elegance, immersion, and coherence. QR codes embedded throughout the issue connect print to video interviews, expanded features, and digital exclusives, collapsing the distance between formats and turning each issue into a connected experience rather than a static product.

At the center of the relaunch was Newsweek’s Newsmakers franchise, anchored by a deeply introspective profile of Sir Anthony Hopkins, paired with a high-production video interview and a collectible print cover. It is a quiet statement of intent. Print legacy is no longer the destination. It may still be an on-ramp for authority.

This is why Newsweek is the right subject for a serious examination of media in the age of agents. It is old enough to remember what authority once was. And disciplined enough to understand what authority has become.

The rest of this essay is an attempt to explain that transition, and what it teaches us about the future of media itself.

Analysis: Why, Newsweek?

There was a time when the covers of Newsweek could tilt the emotional posture of a country.

You could feel it when one landed: airports, hotel lobbies, and kitchen tables. The magazine did not chase the week; it decided what the week was about. That power was not just cultural. It was infrastructural. Information moved slowly. Distribution was physical. Attention was concentrated; uthority was something you earned once and then exercised repeatedly. That system collapsed quietly.

Newsweek is not failing at being what it was. It is succeeding at being what the system now demands. And the system is only becoming more agentic.

The internet did not destroy journalism. It destroyed the economic and mechanical conditions that made that kind of authority possible. What replaced it was not chaos but a new structure, one built on velocity, redundancy, and continuous recomposition. And now, layered on top of that, a second transformation is underway. Agents increasingly decide what information matters before any human ever sees it.

Media brands are no longer competing primarily for readers. They are competing for position inside automated systems of discovery and synthesis. In this environment, there are three viable ways to run a modern media organization.

Some lean on brand equity and opinion driven journalism; their product is identity. They gather loyal audiences around recognizable voices and coherent worldviews. This remains extraordinarily effective for shaping human belief. It is far less effective for shaping machine behavior.

Others pursue channel innovation and deep reporting. They break stories and then, they generate primary facts. They still perform the essential labor of journalism. But their work moves slowly and expensively, and the distribution of that work increasingly depends on systems they do not control.

A third group has recognized that authority itself is changing form. These organizations build on brand equity, but they optimize relentlessly for the way agents ingest, rank, and reuse information. Their journalism is designed less to persuade and more to persist.

This is where Newsweek now lives.

The modern Newsweek is not the nostalgic magazine that people remember. It is not built to freeze a moment in time or to shape national sentiment with a single image and a handful of words. It is built to remain permanently inside the bloodstream of the global information system.

That transition did not happen because of ideology; it happened because of math. Consider this simple back and forth between a friend and me:

I’m starting to notice this trend across multiple media outlets. Scared of being sued? Has Trump drawn new lines? Have news firms grown a conscience? Are people simply becoming wiser to inductive framings? Either way, it’s a welcome new trend and should reduce the number of people propagating fiction dressed as fact.

My response:

GPT indexing incentivizes factual reporting rather than opinion.

Search engines, aggregators, social platforms, and now large language models reward coverage density, topical breadth, speed, and recognizability. They favor brands that publish often, across many domains, and with framing that is legible to both humans and machines. They reward familiarity. They penalize obscurity. They quietly assign influence based on retrievability and perceived risk. They also consider objectivity and seem to weight the reporting that abides by it.

Newsweek adapted to that reality.

It publishes at scale. It covers everything from geopolitics to culture to technology with high velocity. It rarely attempts to own the first narrative. Instead, it enters the stream quickly, frames events in language that travels easily, and becomes one of the sources that other systems feel safe citing.

This is not an accident. It is a strategy.

In the agentic world, credibility is no longer primarily something humans feel. It is something models compute. The Newsweek masthead still carries decades of accumulated trust. That trust now functions as a kind of passport inside automated knowledge systems. As suchg, it lowers friction. It reduces perceived liability. It increases the probability that an agent will include Newsweek in its summary of reality.

This is why Newsweek’s longevity matters deeply in the age of agents, even as its legacy no longer defines its business model. The nostalgia attached to old covers belongs to a different economy of attention. The modern Newsweek does not sell time. It sells placement.

There is something unsettling about that shift, especially for those who grew up with a very different conception of what journalism was supposed to do. The new model does not optimize for persuasion or for revelation. It optimizes for endurance inside a system that never stops recomposing the world. But it is a mistake to read this as decline.

Newsweek is not failing at being what it was. It is succeeding at being what the system now demands. And the system is only becoming more agentic.

The deeper implication is uncomfortable. In the coming decade, media power will belong less to those who write the most beautiful stories and more to those who fit most cleanly inside the machinery that decides which stories exist at all.

That machinery is already here. Newsweek simply learned how to live inside it. That is the real story: the internet used to be direct-to-consumer; today it is a wholesale retailer. Agentic stocks our attention economy’s shelves with commodities; the consumer no longer has to discover or concern itself with the brand behind the message. The arbiter of truth sources its news from elsewhere; we await its interpretation.

作者：Web Smith

Agentic: The New Demand Layer of The Internet

A 2PM Hypothesis by Web Smith

The consumer internet is reorganizing itself around a new gravitational center, one that few operators predicted and even fewer are prepared to serve. For twenty-five years, the web revolved around search boxes, category trees, filters, and social discovery. But the moment autonomous agents began negotiating, filtering, and shaping demand on behalf of consumers, the ground shifted. What started as a novelty: chatbots answering trivial product questions—has evolved into a structural redirect of the consumer funnel. Agents are no longer assistants. They are becoming the new point of sale, the new homepage, the new SEO layer, and the most trusted advisor in a digital environment drowning in information.

This transition is not theoretical. It shows up in the math. Chatbot-to-eCommerce referral traffic grew 1,300% in 2024 and another 520% in 2025, driven in large part by the rise of the shopping-agent layer that OpenAI introduced quietly, then relentlessly optimized. The early evidence suggests that agentic eCommerce is not simply an efficiency tool; it is a reorganization of intent itself. The brands that understand this early are not competing for marginal gains in impressions or clicks; they are preparing to dominate the next decade of distribution.

This is the context in which Agentic Commerce takes shape, elevated by the discipline we now call Agentic Engine Optimization (or AEO). AEO succeeds the alphabet soup of the last era (SEO, CRO, growth hacking) by asking a more fundamental question. Instead of “How do I rank on Google?” the operative question becomes: How does my brand win the recommendation when an agent chooses on behalf of the consumer?

I believe that I have figured out how to achieve this on behalf of brands.

Consumers are overwhelmed. They want fewer decisions, lower cognitive load, and far less time spent sifting through 4,000 Amazon SKUs in search of a product that fits an immediate need. They want trust. They want relevance. They want clarity in the exact moment a problem presents itself. AI agents are uniquely equipped to deliver this. They compress research, comparison, and validation into a single intent-driven interaction. They ask clarifying questions, reconcile contradictory preferences, prune irrelevant options, and return a curated set built from verified product data. They search laterally across retailer catalogs: Target, Walmart, eBay, Etsy, Shopify, and independent brands—aggregating selection in a way that even Amazon cannot fully replicate.

The traditional discovery stack is collapsing under the weight of its own complexity. What emerges in its place is a new linear progression: a prompt, a specification, an agentic query, a curated set, and a transaction. Every brand has to ask itself where it surfaces in that sequence and more importantly, why. That is the work of AEO.

AEO is, at its core, the operational science of making a brand legible and trustworthy to autonomous agents. Modern answer engines are revealing their preferences in real time. Agents gravitate toward structured, verifiable facts: Brand Facts pages, ItemLists, Product schema, policy clarity, comparison matrices, and machine-readable truth surfaces. The “Rank First in ChatGPT” frameworks being developed across the industry map neatly onto this reality, outlining the schemas, structures, and canonical pages that agents parse effortlessly and prefer to cite.

They also reward consistency. When your PDP schema, Shopify feeds, GTINs, product attributes, and delivery windows contradict one another, agents downgrade your credibility. When those elements harmonize, agents elevate your recommendations. Clarity (not creativity) is the currency of this new ecosystem. Plain-language headings outperform clever taglines. Direct answers outperform 700-word stories. Predictable URL structures outperform branded flourish. The DTC era prized narrative; the agentic era prizes truth.

But facts alone do not win the future.

Agents need contextual hooks. They need to know not just what a product is, but when it matters, why it matters, and for whom it matters. This is where your frameworks, your decades-long instinct for physiologic timing, consumer cycles, and context-driven merchandising, become a competitive advantage. Your thesis has always implied a formula: Demand emerges at the intersection of timing, physiology, problem, and availability. AEO simply formalizes that intuition.

No concept demonstrates this more clearly than MTN Haus’ engineering design for the omni-channel CPG “Snack Clock.” (concept development here)

Snack Clock is an early expression of Agentic Commerce theory; the system sounds so elementary that it lacks the appeal of what’s typically viewed as commerce innovation. It is rooted in a simple but profound insight: people do not buy products, they buy solutions that arrive at the right moment. Every product solves a problem that spikes at a particular time of day. The afternoon slump is a time-sensitive problem. The morning electrolyte gap is an easily solvable problem. The post-run stupor period is an easily solvable recovery issue. The pre-dinner craving window is a satiety problem, also solvable. eCommerce has historically ignored time as a demand vector; the Snack Clock concept restores it.

More importantly, it eliminates cognitive load. Instead of browsing a store, the user receives a contextual nudge: “It’s 2:42 PM. Your energy cycle is usually dipping by now. Here’s what you usually need.” This is AEO in motion and eventually, these agents will think for us. As such, the product finds the consumer; the consumer no longer has to find the product. And as agents begin building temporal models (hydration cycles, circadian rhythms, workout patterns) the brands anchored in timing will surface first. Snack Clock becomes the training set that future agents rely on. A decade from now, every major brand will attempt to build its own version. The savviest, most well-resourced will build there’s early.

The broader eCommerce stack is now splitting into two spheres. On one side is the Amazon Sphere: a closed system built on fulfillment dominance, infinite aisle selection, and Prime lock-in. On the other side is the Chatbot Sphere: an open discovery layer where agents evaluate products across retailers, ask qualifying questions, and narrow options for higher-consideration purchases. Every brand outside Amazon’s walls must optimize for the Chatbot Sphere or risk becoming invisible. Agencies like the one where I spend the vast majority of my time (MTN Haus) sits at the center of this shift, uniquely positioned to build the infrastructure required for brands to be agent-ready. Enterprise clients, operational experience, and just enough authority to push back on clients who’ve yet to study these evolving demands.

Most Shopify brands have no idea how to make themselves visible to agents. They assume this future will resemble SEO. It will not. Migration to structured, agent-friendly Shopify systems will increase in demand. Growth Design that blends Baymard-levels of science with Freudian-like behavioral clarity. Content architectures built around comparison hubs, brand facts, policy transparency, and schema alignment will be commonplace. And contextual UX systems, led by Snack Clock architecture, that give agents the signals they need to recommend products at the right moment.

This is not a marketing channel; it is a reorganization of demand itself. The executives who thrive in this era will understand that search is no longer the arbiter of discovery. Agents decide what consumers see first. Brand visibility becomes a technical discipline. Merchandising evolves into decision architecture. Timing and physiology become competitive edges, not afterthoughts. And eCommerce agencies will rise in demand, ones that don’t just deliver sites but also deliver agentic-ready ecosystems capable of being recommended by the systems that now shape demand.

The gap between agent-native brands and legacy brands will widen quickly. The brands that embrace structured facts, contextual timing, schema alignment, and truthful ecosystems will outperform. The brands built on narrative-heavy storytelling, homepage vanity, and inconsistent data will fade. Agents reward clarity and punish noise. The last decade of ecommerce was built on noise; the next will be built on structure.

Every major reorganization of digital commerce follows a familiar arc. Search produced SEO. Social produced organic demand generation. Mobile produced conversion optimization. AI agents are producing AEO and the contextual commerce layer that sits atop the modern web. We are at the threshold of this fourth reorganization. Every brand will need an agentic strategy. Every retailer will need an agent-ready catalog. Every operator will need their version of the “Snack Clock” and the engineering teams capable of building the middleware to enable it.

The brands that treat this moment as a trend will lose ground.

The brands that recognize it as a structural realignment of digital commerce will capture demand their competitors will never see. The winners of the next decade won’t simply be visible. They will be recommended. Agentic Commerce is the new shelf space. AEO is the new route to market. The theories behind “Snack Clock” architecture will be the new CRM. And brands will partner with agencies, like never before, to build the infrastructure that will power all three.

作者：Web Smith