Category: Thinks

NYTimes: “For decades, Americans scoffed at Russia’s rigid, centralized military and its inability to adapt. That picture is dangerously out of date. After four years of war in Ukraine, Moscow has developed an impressive, pragmatic approach to military innovation that prioritizes what works over what is elegant, what scales over what is ambitious, and what delivers battlefield results over what impresses on paper. Russia is reshaping the future of warfare in real time, building artificial intelligence-enabled command and control and, it appears, deploying fully autonomous weapons without the ethical constraints that govern Western militaries.”

Tyler Goodspeed: “There’s a two-step process to identifying historical economic recessions. The first step is quantitative. You must look at the statistical evidence. Was there an ongoing economic contraction? The second step is more qualitative. At the time, what were people concerned about economically? What were they lamenting? What were their economic concerns? That’s different from what they were blaming the recession on. When you look at the statistical evidence—the quantitative evidence—you very often see violent changes in the rate at which people are unemployed—sharp contractions, rather than gentle ones, where you tip or you slip into recession. With the qualitative evidence, you often hear people speak about “big shocks” or big clusters of shocks.”

WSJ: “Nish Ajitsaria, senior managing director, head of Aladdin Product Engineering, and the firm’s executive sponsor for AI, said he is chasing down a future in which AI becomes the default mode for executing most processes, from research to coding. Meanwhile, human roles will become less specialized, and more cross-functional, working in more nimble “squads” to oversee the AI’s busywork.”

Alex Imas: “After trying to streamline the store experience with fewer workers and more automation, [Starbucks] concluded that this had been a mistake. CEO Brian Niccol said that “handwritten notes on cups’’, ceramic cups, and “the return of great seats’’ had led more customers to “sit and stay in our cafes’’, showing that “small details and hospitality drive satisfaction.’’ More baristas are being hired per store and automation is being rolled back…Economics is the study of decision-making under constraints, i.e., scarcity. If advanced AI brings material abundance—if machines can produce many if not all forms of human production at very low marginal cost—does economics become irrelevant? No, we will still have scarcity, but the kind of scarcity that matters will change. Ultimately the answer to any question about the future economics of advanced AI begins with identifying what becomes scarce.”

NYTimes: “[Javier Milei] wants to use his presidency not only to slash the country’s budget but to wage an ideological war and rewire the country’s mentality. He wants to dismantle what he calls the “aberrant” concepts of social justice and economic equality and make the nation’s core principles capitalism, the free market, a limited state and individualism. “We are at war,” Mr. Milei said at a right-wing festival last year, and added: “We are fighting a cultural struggle, an ideological battle, a war for the survival of our freedom.” At political rallies and international summits, in public policies and a deluge of social media posts, Mr. Milei has relentlessly sought to infuse Argentina with his libertarian ideals. And turn it into a model for the world.”

WSJ: “China is making strides in open-source artificial intelligence. Eighty percent of developers worldwide who use open-source AI tools are building with Chinese models, according to an estimate by our colleague Martin Casado, general partner at Andreessen Horowitz. Research from our firm and OpenRouter shows a significant increase in the use of Chinese open models last year, reaching in some weeks a high of 30% of all AI usage. In January, Alibaba’s Qwen family surpassed 700 million downloads to become the most widely adopted open-source AI system on the planet.”

Derek Thompson: “In thinking about the right historical analogy for AI, I’ve become very interested in the early history of electricity in the early 1900s.”

David Henderson: “Four facts about taxes: 1. High marginal tax rates cause economic harm. 2. High tax rates also cause tax avoidance. 3. Making tax rates the same for everyone would likely reduce the demand for government spending. 4. Most people, not just high-income people, think a proportional tax on income is more fair than a graduated tax with higher rates for higher-income people.”

WSJ reviews “The Revolutionary Center”: “What to make of his plea that liberalism needs a “great rebalancing”? Mr. [Adrian] Wooldridge spells it out: Dial back the elitism; decentralize democracy; rid liberalism of its recent taste for extremism; make left-liberals value excellence again; make neoliberals get a conscience; and, most intriguingly, make liberalism more pessimistic, less “drunk on hope.” “The most pressing liberal task,” Mr. Wooldridge writes, “is not to create a utopia. It is to prevent terrible things from happening.” Prudence is in short supply, and it is this very scarce resource that requires us “to block the roads to hell before building the stairways to heaven.” Liberals of the world unite—but curb your ambition.”

FT: “Implicit in this work is the assumption that, one day soon, the world’s robot population will explode and that we are not prepared. If, as many are predicting, that number grows from millions into billions, where will we find the resources to build them? Who will take care of them? Where will they go when they die? The creative machinists want to build robots that answer these questions themselves. Robots that “eat” and “heal” and “reproduce”, and what is reproduction but the ultimate form of self-repair? One day, Wyder told me, we’ll purchase not a robot but a bagful of robot, filled with building blocks. The blocks will assemble themselves to take whatever shape and perform whatever task.”

Ben Thompson: “My bet is that owning demand will ultimately trump owning supply, suggesting that the underlying principles of Aggregation Theory lives on. To put it another way, I think that OpenAI will need to win with better products, not just more compute; then again, if more compute is the key to better products, then does supply matter most? Regardless, they’ll certainly be focused on delivering both to the enterprise customers who are driving Anthropic’s astonishing growth. The real cost may be the consumer market they currently dominate, given that Meta has nothing to lose and everything to gain.”

SaaStr: “If you’re a B2B vendor and you can’t ship an AI feature that is genuinely best-in-class … don’t ship it. Because your best customers, the ones who would actually pay for it, are going to try it once, see the 0% sentiment score and the empty recommendations tab, and decide your platform “doesn’t really do AI.” And then they’ll buy the AI point solution.  This could potentially all change down the road.  But not today.  Today, you lose with a 60% AI Solution.”

WSJ on retirement: ““For people who’ve had careers that have given them both an identity and status—and consumed most of their energy and attention—stopping work can feel like stopping everything,” says Ruth Finkelstein, executive director of the Brookdale Center for Healthy Aging at Hunter College in New York. “It takes a recovery process to be able to enter a chapter that’s defined differently.” Such a process often begins with a plan; after all, if career and ambition are central to your existence, replacing them with nothing isn’t going to work.”

SaaStr on Forward Deployed Engineers: “FDEs sit at the intersection of product, engineering, and customer success. They go on-site or deep into a customer environment, understand the actual workflows, and configure the product to work inside those workflows. They’re not building from scratch. They’re not doing basic support. They’re doing the hard middle work of making software actually land in the real world. This role has existed for a long time. Palantir essentially built their entire go-to-market around it. You couldn’t buy Palantir’s product and self-serve your way to value. You needed their people inside your organization, configuring and training the system for your specific context. That model worked. It was expensive and it didn’t scale the way SaaS was supposed to scale. But it worked, because complex software in complex environments requires human judgment to deploy well. Now almost every serious AI product has the same requirement. And almost no one has enough of the people who can do it.”

Jensen Huang: “In the end, something has to transform electrons to tokens. The transformation of electrons to tokens and making those tokens more valuable over time is hard to completely commoditize. The transformation from electrons to tokens is such an incredible journey. Making that token is like making one molecule more valuable than another molecule, making one token more valuable than another. The amount of artistry, engineering, science, and invention that goes into making that token valuable, obviously we’re watching it happen in real time. The transformation, the manufacturing, all of the science that goes in there is far from deeply understood and the journey is far from over. I doubt that it will happen.”

NYTimes: “The strange way that A.I. looks like a genius at one moment and dense in another is what researchers, engineers and economists call “jagged intelligence.” They use this term to explain why A.I is racing ahead in some areas — like math and computer programming — while still struggling to make headway in others. The term, which is widely used by the people building A.I. and analyzing its effects, could help reframe the debate over whether these systems are becoming as smart as, or even smarter than, humans. Instead, researchers argue, A.I. is something completely different: far better than humans at some tasks and far worse at others.”

Alexis Krivkovich: “The great paradox is just that. Companies expect massive transformations from AI and have invested with that mindset. Yet more than 80 percent of companies say they’re not yet seeing impact on the bottom line from those investments. The real question is “How do we meet this moment and create the agentic organization of the future?” In every conversation I have, leaders feel like they’re on the precipice of that question. What they’re trying to think through is really a set of questions: “How are roles going to change? What are the skills we’ll need in the future? How do I bring along our workforce with excitement, not fear? How do I drive that change so it hits every corner of the organization?””

Morgan Housel: “It’s the classic John Adams line, which I’ll paraphrase: “I studied war so my kids will have the liberty to study engineering. They will study engineering so their kids can have the liberty to study philosophy, whose kids can have the liberty to study art.” I hope my kids and grandkids won’t have to worry about cancer in the ways we do. I hope they have incredible technology that makes their jobs easier than ours. I hope that everyday frictions we deal with today disappear. I hope their energy is so abundant they consider it unlimited. Is that spoiled? I suppose, but when you frame it like that you might think of a different word – perhaps “lucky,” or, “fortunate.” Or perhaps, “beneficiaries of the accumulated hard work of those who came before them in a way that leaves them able to spend their days solving new problems.” Which is what you and I are today.”

WSJ: ““Everyone’s talking about oil, but I think what the world is mainly short of is tokens,” said Ben Pouladian, an engineer and tech investor based in Los Angeles. A token is a unit of measurement in AI to track how much computing resources are being used for a task. “AI is at this point no longer just some chatbot that we ask for a recipe while we stand in front of the fridge. It’s orchestrating tasks, it’s getting smarter,” Pouladian said…Historically, price increases have been among the only ways to address a supply crunch, but such a move could be perilous for frontier AI companies, which are in a ferocious competition to gain users. Hourly rental prices for GPUs, the microchips used to train and run AI models, have surged since the fall. Anthropic, the maker of popular chatbot Claude and viral coding app Claude Code, has been plagued recently by frequent outages. The company has begun metering computing supply to users during peak hours, but the rollout has been marred by customers who have complained that they are reaching the limit far too quickly.”

NYTimes: “India’s investments in renewable energy, especially solar, have been staggering. By the end of last year, it had 55 solar parks, including one that stretched across 14,000 acres of desert. The solar farms alone were producing 40 gigawatts — enough energy to power about 80 million rural households. Smaller rooftop arrays sprouted seemingly everywhere. India announced last summer that it could meet 50 percent of its electricity requirements with renewables alone, years ahead of schedule. But solar and wind power don’t flow on demand. India’s critical challenge is storing surplus energy from peak production and delivering it consistently when people need it to run businesses or turn on a stove. Battery systems and transmission lines can’t be built fast enough.”

NeoMails and NeoNet are the Building Blocks – 1

How do you actually build a media network on relationship attention?

The answer has three layers.

NeoMails create the attention surface. ActionAds monetise it. NeoNet makes it portable across brands.

Together, they form what I think is the right name for this new category: the Inbox Media Network.

The attention surface: NeoMails

NeoMails exist because conventional email failed to build a third message class. Brands had Sell and Notify. They did not have Relate. NeoMails are built specifically for that missing function: creating a lightweight, repeatable, worthwhile interaction between transactions.

The APU — BrandBlock, Magnet, Mu, ActionAd — is the atomic unit of that design.

The BrandBlock gives the interaction brand identity and context. The Magnet earns participation: a quiz, a prediction, a poll, a preference fork — completed in under sixty seconds. Mu gives continuity and visible memory — an attention currency that accumulates with each interaction, visible in the subject line, creating the habit loop that sustains daily return. And the ActionAd funds the send.

This architecture means the inbox is no longer being monetised only when the customer is buying. It is being activated and monetised when the customer is simply paying attention. A customer opens a NeoMail, completes a Magnet, earns Mu, and sees an ActionAd — all in sixty seconds, all inside the inbox, all generating a first-party signal for the brand. The relationship layer, which was previously invisible to the marketing P&L, now has an economic address.

The parallel with habit-forming consumer products is instructive. Products that achieve daily engagement do not do so by making each interaction intensive — they do it by making each interaction short, rewarding, and consistent. The Magnet is the micro-interaction. Mu is the streak. The habit that forms is not the completion of a task. It is the maintenance of a relationship.

The monetisation layer: ActionAds

ActionAds are not banners placed inside an email. They are in-email action units that complete inside the inbox, on authenticated identity, at the moment of peak attention. The advertiser pays for the action, not the impression. The sending brand earns revenue from a customer who may not be buying today but is still engaged enough to act.

Two formats power the Inbox Media Network.

The One-Tap Subscribe ActionAd subscribes a customer to another brand’s NeoMails with a single tap. The email address is pre-filled. No landing page. No form. No confirmation loop. Explicit consent logged at the moment of interaction. The subscribing brand pays a transfer fee per confirmed, consented, NeoMail-active subscriber — a fraction of what the same customer would cost through a paid media auction.

The form-fill ActionAd generates a verified lead inside the inbox — contact details, a preference, a qualification question — completed without the customer leaving the message. The advertiser pays a cost-per-lead fee, split between Atrium and the publishing brand.

That split is what closes the ZeroCPM loop. Instead of asking a brand to fund a Relate message as a cost centre, the system allows the message to fund itself. This is not a small accounting detail. It is the inversion that makes relationship attention economically rational for brands that would otherwise never invest in it.

This also resolves the incrementality tension directly. In the old model, the question is whether the email drove the conversion or merely arrived in the presence of a decision that had already been made. In the NeoMail model, the value of the message does not depend on forcing a discount-led conversion. It creates attention, action, and monetisation even before any purchase happens. The channel becomes less dependent on contested attribution because it now has a direct economic layer of its own.

NYTimes: “Everyone knows the feeling of being stuck on hold, or spending time clicking through an endless series of pages to cancel a subscription, or screening spam calls, or trying to change a flight. These inconveniences make life less pleasant, but they are easy enough to write off as small concerns, as the mere stuff of living a contemporary existence. It turns out that such time-sucking, tedious tasks have a meaningful economic impact. The accumulated cost of what some refer to as the “annoyance economy” adds up to $165 billion a year in lost time and wasted money for American families, according to a new report from Neale Mahoney, a Stanford economist, and Chad Maisel, a policy fellow at Groundwork Collaborative, a progressive research organization. The annoyance economy includes “the everyday interactions that should be simple but often turn into fraught ordeals,” Mr. Mahoney said.”

Bloomberg: “India’s $315 billion IT industry is one of the main engines of the country’s economy. For decades, coveted jobs at Infosys and competitors HCL Technologies, Tata Consultancy Services (TCS) and Wipro offered millions of college-educated Indians a path into the upper middle class. The work involved building software systems for foreign airlines, Wall Street banks and Silicon Valley giants. But these stalwarts are caught up in an existential crisis. Rapid advances in artificial intelligence have called into question whether software companies even have a future. The release of specialized versions of Anthropic PBC’s Claude AI assistant and autonomous agentic AI tools in February unleashed a global “software-mageddon” that wiped out about $800 billion in stock value in a single week. India’s Nifty IT index plunged almost 20% for the month, its biggest drop since the 2008 financial crisis.”

WSJ: “Welcome to the era of the mega-layoff. In Silicon Valley and beyond, companies that are cutting staff are doing it with a big ax. Instead of laying off people in more incremental—and less disruptive—waves, employers are seizing on the potential financial upsides of severing swaths of their workforces at once. That is a departure from not long ago, when mass layoffs registered as a sign of trouble or mismanagement and that a company needed to take drastic measures to right its performance. Now, such a company is more likely to get a big stock bump and praise from investors for acting boldly.”

NYTimes: “China, the U.S., Russia and others have ramped up their contest over artificial-intelligence-backed weapons and military systems. The buildup has been compared to the dawn of the nuclear weapons age.”

Why the Inbox Is the Most Underleveraged Attention Surface in Marketing – 2

Property four: portable and permanent

Social accounts get deleted. Apps get uninstalled. Device IDs break. Cookies expire. Platform relationships depend on the platform remaining relevant.

An email address endures. The same address works across every device, every platform, every application — independent of any single ecosystem’s fate. It is the most portable identity credential in digital marketing, and it has been for thirty years. The brand that has maintained an inbox relationship with a customer for five years owns something that cannot be replicated by any amount of paid media spend.

Why it has been underleveraged

Given these four properties, the question is not why the inbox should become a media network. The question is why it took so long.

The answer is that brands built the wrong product on top of the right surface.

Most email programmes reduced the inbox to two message classes: Sell and Notify. Sell messages drive the next transaction. Notify messages confirm what just happened. Both matter. Neither is enough.

The problem is not that these messages are bad. The problem is that they leave the long middle of the relationship empty. When the customer is not in-market — which is most of the time — the brand has little to say except more extraction. The inbox becomes a broadcast pipe: launch, offer, cart, urgency, receipt, alert, repeat.

That is what degrades the surface.

Once the inbox is treated only as a vehicle for asks and confirmations, its relationship value collapses. Open rates fall. Relevance gets measured only through conversion. The customer either buys or drifts. And when they drift, the same brand pays to reacquire them through paid media. The already-committed customer keeps receiving discount offers they did not need, until they expect a discount before every purchase. The channel looks healthy. The margin tells a different story.

The right argument for the inbox is not a channel case. It is an attention-surface case.

The surprising claim is not “email still works.”

The surprising claim is this: the inbox is the most valuable underleveraged relationship-attention surface in marketing. And the reason it has been underleveraged is not a technology limitation. It is a product limitation.

That is where AMP for Email changes the equation. Most marketers still think of email as a document. AMP changes its category. A message can now accept input, process forms, update content, and complete actions inside the inbox. The inbox stops being static. It becomes a surface.

That shift is fundamental. Because a media network requires action. A relationship surface without action is content. A relationship surface with action can become infrastructure.

The regulatory tailwind

The timing of this shift is not accidental. Four forces are converging simultaneously, and all four point in the same direction.

Cookie deprecation is removing the foundation of most programmatic advertising. Privacy regulation — GDPR, national data protection frameworks, and the broader global tightening of third-party data — is restricting the alternatives. AI is flooding every rented platform with more content and more competition, making purchased attention noisier and more expensive. And brands are beginning to understand — through incrementality analysis, through CFO scrutiny of ROAS, through the realisation that much of their attributed email revenue was correlation rather than causation — that the reporting they have relied on is inflated and that the reacquisition spend hiding inside their acquisition budgets is structural.

These are not headwinds for the inbox. They are tailwinds. Every regulatory constraint on third-party data makes authenticated first-party identity scarcer and more valuable. The inbox does not need to become the future. It needs to be recognised properly in the present.

The inbox already has the ingredients.

What has been missing is the machinery.

Martin Casado: Assume there’s 30mn professional software developers or something like that, and they make an average of $100,000 a year. This is a $3tn market. I mean, we’re just in the very, very early innings of this massive market.  A single lawyer can now do more cases, the accuracy of healthcare will increase. We know that the call centre stuff is working. We know customer support is working. But then there’s a long tail of stuff that we’re not quite sure about. It may also be the case that AI makes us feel more productive than we actually are. That’s a kind of phantom productivity. How often would you talk to ChatGPT, and it’s like, ‘that was such a great idea.’”

FT: “The dollar still accounts for roughly 57 per cent of global foreign exchange reserves and international payments, and dominates other aspects of international finance. Capital inflows into dollar assets, including US Treasuries, remain strong. True, the premium on Treasuries relative to other countries’ government bonds has declined, but that on other risk-free US dollar assets has not.”

Zvi Mowshowitz: “One of the stronger arguments against further AI progress was that scaling the models had stopped working. We had a standard ‘full-size’ for models like Gemini 3.1 Pro, GPT-5.4 and Claude Opus 4.6. If you wanted a better answer, you had it think smarter and for longer, and in parallel, but you didn’t scale it bigger because that wasn’t worthwhile. Now we see that this is not true. It is worthwhile again. That changes things a lot, and in terms of existential risks and related concerns it is not good news. What to do about it? Unprompted, same as always, various people say ‘this only means we need to move forward, because if we don’t someone else will.’ Well, sure, they will with that attitude. Pick up the phone. Get to work. Lay the foundation.”

WSJ: “[Edward] Fishman says a chokepoint has three attributes. First, a country or coalition must have a dominant enough market position to swing the supply or price of a commodity or service. Second, substitutes for that commodity or service must be hard to find in the short run (in the long run, everything has a substitute). Third, closing that chokepoint must have asymmetric effects, i.e., hurt your adversary more than it hurts you.”

NYTimes: “For decades, two kinds of scarcity kept the internet safe — or safe enough. Writing software was hard, so the people who did it were trained, careful and few. Finding bugs was also hard, so the worst flaws stayed hidden, sometimes for decades. It wasn’t a great system. But the difficulty on both sides created a kind of détente that held. Now, thanks to new A.I. tools, anyone can write code. Soon, bad actors could use those same tools to find out what’s wrong with code. The détente is over.”

Akash Prakash on the FPI narrative on India: “While we are probably at peak negative India sentiment, and this sentiment has some element of cyclicality, with the AI trade front-loading earnings traction in North Asia, it is not entirely wrong. We need to regain the growth narrative. Today most investors see no compelling reason to look at India. Not cheap enough, not growing fast enough, not a global leader in any technology, why be here?”

FT: “Can humans still flourish in a near-future jobs market dominated by AI? Yes we can, says the theoretical neuroscientist and social entrepreneur Vivienne Ming. But first, she argues in her new book Robot-Proof, we must change the way we educate and hire the next generation of students and graduates. Equal parts self-help book and corporate manual, Robot-Proof is aimed at the home and the C-suite, its arguments dispatched with a Bay Area confidence that its readers inhabit both (in one section Ming advises parents to move “from a model of knowledge transmission to one of capacity building”). Will the world we’re preparing our children for exist by the time they have completed their education? No, argues Ming: AI will “deprofessionalize” once high-skill careers, opening up the law, medicine and finance to lower-skill, lower-paid workers and transforming the day-to-day of those jobs into something resembling a production line.”

Stanford’s 2026 AI Index report. “AI capability is not plateauing. It is accelerating and reaching more people than ever. Industry produced over 90% of notable frontier models in 2025, and several of those models now meet or exceed human baselines on PhD-level science questions, multimodal reasoning, and competition mathematics. On a key coding benchmark—SWE-bench Verified—performance rose from 60% to near 100% in a single year. Organizational adoption reached 88%, and 4 in 5 university students now use generative AI.”