Author: Rajesh Jain

An Entrepreneur based in Mumbai, India.

Thinks 2013

FT: “Today, quantum computers are no longer just found in research laboratories as companies explore the technology’s commercial possibilities. Industry observers estimate there are scores of quantum computing systems in the world, a number forecast by the consultant McKinsey to rise to around 5,000 by 2030…As the new machines begin to indicate they can outperform conventional computers in niche areas — what the industry dubs “quantum advantage” — the risk of dismissing the technology’s importance is growing. Organisations are readying themselves for the prospect of Q-Day — the predicted moment when quantum computers are capable of breaking the cryptographic methods on which modern societies rely.”

WSJ: “Oura and Whoop do have advantages. Rather than selling mass-market step counters, they market premium products that synthesize biometrics like heart rate variability and skin temperature into actionable outputs: early illness warnings, recovery scores and training recommendations. Both benefit from sticky subscription revenue and, notably, lack screens. That positions them both as companions to an Apple Watch or as alternatives for users who don’t want yet another screen to look at.”

Mint: “A major reason why I consider small chat an art is that many leaders in the corporate sector seem much too shy. They are trained to face the wrath of shareholders and a stern board but perhaps feel self-conscious if they spot a group of executives in the corridor. While walking past without pausing to chat may be tempting, it could appear rude. Perhaps they fear letting out top-level confidential information. Or maybe they do not see it worthy of their time, which may explain the hems and haws that precede an excuse to slip away. Lost opportunities only pile up this way. The art of small talk, which diplomats and dignitaries have mastered, must get its due credit in the corporate world as well. This form of engagement is an important part of the learning curve of leadership. Even a tiny exchange could have a powerful impact.”

Sarah Guo: “As Gabe Pereyra says, real automation isn’t only the model getting better. It’s the product, the model, the workflow, and the firm moving together, and three of those four move at the speed of an organization. Moving people is the part no benchmark touches: getting a skeptical partner to change how she runs her matters, holding a team together through a rebuild. It’s why, when we hire a CEO, the ability to deal with people weighs at least as much as the analytical horsepower, and a smarter model doesn’t change that weighting. The feedback is ambiguous, the horizon is years, and the trust belongs to a person. Every company I know has every engineer on frontier coding models, and not one has changed its eng org at anything close to that speed. Adoption took a quarter, and what a magical quarter of token growth it was! But the rebuild is taking years…What I’d bet on is the direction: intelligence keeps getting cheaper, and value keeps sliding toward the few places a model can’t reach. The untrainable is value with history.”

Thinks 2012

Ruchir Sharma: “America’s profit machine seems extraordinary by historical and global standards. But look closer, and cracks appear. Rising government deficits explain a surprising share of recent US earnings growth. Moreover, the “profitless” dotcom era is a myth. Earnings growth is not dramatically stronger today than it was in the late 1990s. Since then, speculative excess has moved into private markets, making the public markets and the economy look more robust than they really are. In short, this expansion is more dependent on government and the earnings story is less exceptional than investors realise. Overall corporate earnings have risen from 7 per cent of GDP in the late 1990s to 11 per cent today. The dynamism of American business has played a role, but so have tax cuts and government spending. Lately the US deficit has risen to more than 6 per cent of GDP and a deficit that high reflects a large transfer of income to households and corporations.”

CNBC: “There are 857 U.S. startups valued at $1 billion or more, the threshold for being deemed a “unicorn” company, according to PitchBook data. But nearly half of that group hadn’t raised fresh funding in the last three years as of the end 2025, making many of those valuations stale, according to the private markets data firm. Startups that last raised in 2021 were worth 68% less on average at the end of last year, while those that last raised in 2022 saw a 52% decline, according to Pitchbook’s own valuation estimates. As a result, more than 220 companies that had reached billion-dollar valuations in the venture boom were deemed fallen unicorns, according to PitchBook, which provided a list of the companies exclusively to CNBC.”

NYTimes: “The reading crisis is real. But we don’t need new inventions to build a reading city. Exempt books from sales taxes the way we exempt prescription medicine. Invest in library collections and reduce wait lists for books. Open nonprofit and hybrid bookstores when the market alone cannot sustain them. Build on the models that already work: reading in laundromats; libraries in transit systems; books in barbershops, classrooms, homes and pediatric offices.”

TheMaxSource: “Performance marketing is a faucet. Turn it on, water flows. Turn it off, it stops. Growth marketing is a well. It takes longer to dig, but once it’s there, the water costs almost nothing to draw. B2B companies need both. What kills them is running the faucet before the well is dug — or digging a well indefinitely while dying of thirst.”

The Intent You Rent Back (Part 9)

Arun and Maya.

Arun bought his first bag of coffee from Kettl on a Sunday in March, after an ad found him while he was reading about pour-over technique. The beans were good. The unboxing was lovely. He got an order confirmation, a “how did we do?” note three days later, and then, over the following weeks, a steady drip of promotions — 15% off, a new roast, a flash sale. He opened the first two. After that he let them pile up unread. Nothing was wrong. He simply had coffee, a busy job, and an inbox that asked for nothing back.

To Kettl’s systems, Arun was now fading. Sixty days, no opens. The CRM moved him quietly from “engaged” towards “lapsed.”

But Arun had not lapsed. Around day seventy-five the bag ran low, and one morning, standing in his kitchen, he thought: I should reorder. He did not open Kettl’s last email to do it. He typed “best filter coffee” into Google. He scrolled Instagram and saw two rival roasters. He read a comparison on a coffee blog. For about a week, Arun was the most valuable thing in his category — a proven buyer, back in the market, ready to spend.

None of that week happened anywhere Kettl could see. His intent had left the building. It was loud on Google, on social, on a marketplace, and silent in the one place Maya was looking.

Maya is Kettl’s CMO, and she is good at her job. On her dashboard, Arun was an inactive customer: no opens, no clicks, sixty-plus days quiet. Her playbook had two moves for that. The first was a win-back email with a discount, which Arun would not open, because he had stopped opening. The second was the one that worked: load the lapsed list into Meta and Google, and bid. A few days later an ad found Arun mid-search. He clicked, he reordered, and Maya’s report logged a tidy paid conversion. Performance marketing, doing its job.

What the report did not say was that Maya had just paid a thirty-per-cent premium to win back a customer she already owned. She had Arun’s name, his email, his order history, the date of his last bag, and a fair guess at when the next one would run empty. She handed all of that context to the auction and bought back a thin, anonymous slice of it — a probable stranger who was probably Arun. She booked it as a win. It was a tax, and she was paying it on her own customer.

The change, when it came, was not a new tool. It was a different question. Maya stopped asking her team how to improve ROAS on the win-back campaign and asked instead: why are we paying ROAS to win Arun back at all? He was never lost. He went quiet in her channels and stayed loud in the market, and she had simply not been watching the right signal.

So she rebuilt the order of things. First, she gave Arun a reason to keep the inbox open — not a sale, but a short, useful note every couple of weeks: a brewing tip, a single-origin story, a small thing worth thirty seconds. Attention earned, not rented. Second, she let the brand do the arithmetic it already could: Arun’s bag lasts about seventy-five days, so on day sixty-five Kettl reached him first, with a quiet “running low?” before he ever thought to Google. And for the customers who had gone fully dark, where no owned message landed, she found a way to reach them through a trusted partner’s surface before falling back to the auction. She had, without using the words, rebuilt the order from the ground up — own the attention, predict the moment, share through a partner, and rent from adtech only what was truly left.

The next cycle, Arun did not go searching. The reorder nudge arrived the week he started thinking about it, from a brand whose emails he had, lately, started opening again. He reordered in two taps. It cost Kettl almost nothing.

Maya’s dashboard changed too. It stopped showing her lapsed customers after the revenue had already stopped, and started showing her attention slipping while there was still time to do something cheap about it. She could see the drift now — the early fade, weeks before the gap — which is the only window in which keeping a customer costs less than buying him back.

Arun never knew any of this. He just felt, vaguely, that his coffee brand had got better at turning up when he needed it. Which was the point. He had always been Maya’s customer. She had simply stopped renting back the one thing she had owned all along: his attention, and the intent it carried.

Thinks 2011

FT: “What does it take to qualify as a great power? [Brendan] Simms, professor of the history of international relations at the University of Cambridge, argues that four crucial qualities are needed. Does the country have the necessary resources — military, economic and human? Does it have reach beyond the country’s own region and a sphere of influence of its own? Is it seen by its own population and by other countries as a great power — in other words does it have the necessary reputation? And does it have the resilience to absorb losses and to rebound after defeats? Simms uses this framework to explain, in a series of neat potted histories, why second-tier powers don’t make the cut. Japan and Germany, held back by their past and lacking military resources; India, not yet able to project power and a junior player to China in Asia; France, bursting with aspiration but its reach is narrowing and it is too prone to domestic turbulence. Africa and Latin America have regional powers but none qualify as “great”.”

NYTimes: “India is in the throes of Shivaji fever. Across the country, hundreds of statues of the king — usually on horseback, brandishing a sword — have begun studding the broader landscape, popping up in the country’s port cities and along its disputed borders with China and Pakistan. Such tributes to the king, a staple of school history textbooks, had previously been mostly found in Maharashtra, the Indian state dominated by the Maratha community — a broad grouping of Hindus that Shivaji was born into, and that includes farmers and warriors, some of whom are considered lower caste.”

Sam Altman: “I think that was the single biggest driver at what people, of what people realized was, the coding models have so transformed how companies are doing their work and the efficiency and the speed of which they’re able to build products. The coding models got really good late last year, early this year, and then another step forward in recent months. So I think you’re right that that’s the single biggest driver. But we are now seeing scientists really use these models and a much broader application of knowledge work beyond coding. But, yes, I think it’s fair to say that coding is the magic right now.”

FT: “The world’s top AI companies are devoting growing resources to a question many still regard as science fiction: what happens if AI becomes conscious? Google DeepMind, Anthropic and Meta have hired experts in psychology, ethics and philosophy in recent months as they expand research into machine consciousness and AI welfare, according to several people familiar with the matter. The efforts reflect a broader shift inside leading AI labs, where rapid advances in increasingly autonomous systems have revived questions about whether machines could one day possess subjective experience — and what obligations humans might have towards them if they do.”

The Intent You Rent Back (Part 8)

The honest limit, and the new question.

A doctrine is only trustworthy if it states its own limits, so here is NeoNet’s: it does not eliminate adtech. There will always be a residual case — the customer who is in-market but engaged nowhere in the network, searching only on Google, browsing only on Amazon, comparing prices on a marketplace where the brand has no owned or partner visibility. For that genuinely off-network intent, adtech remains useful and is sometimes the only available signal. Paying the tax, for that slice, is rational. The honest test is whether the customer is reachable any cheaper way at all; if not, the auction has earned the spend.

So the doctrine is not never use adtech. It is narrower and far harder to argue with: do not use adtech first for customers you already know. Use owned attention for engaged customers, intelligence for predictable windows, NeoNet for silent customers active elsewhere — and then adtech as the final rung, for the intent nothing cheaper can reach. Adtech is a last-mile detector, not a default one. Used in that position it is a sensible tool; used as the first reflex it is a tax on knowledge you already had.

The question that replaces ROAS

This changes the question a marketing team asks. For years the reflex has been: how do we improve ROAS? The better question, customer by customer, is: why are we paying ROAS at all for this one? If the customer is genuinely new, paid acquisition may be justified. If they are unknown or truly unreachable, adtech may be the only route. But if the customer is already identified, historically valuable and merely silent, the first job is not to optimise the auction — it is to avoid the auction. A brand that asks the second question will find, on inspection, that a large share of its performance spend is being paid to re-buy customers sitting in its own database, drifting through the weakening cells of its own Transaction–Attention Table while no one was watching. That single question, asked honestly of each customer, redraws a media budget faster than any optimisation ever will.

The rule of resort

The whole series reduces to a single rule. Own the intent where you can. Predict it where you can model it. Share it where a partner can route it. Rent it only for what is truly left. Build the Intent Stack in that order and the auction stops being the engine of repeat business and becomes what it should always have been: the last rung, reached for rarely and deliberately. The customer was yours. The relationship was yours. The data was yours. The next intent signal should be yours too. None of this is anti-adtech; it is adtech put back in its place — the last rung, not the first reflex.

Own it where you can. Predict it where you can model it. Share it where a partner can route it. Rent only what is left.

Thinks 2010

Manish Sabharwal: “An uneducated Indian is not a free Indian. Coaching factories, exam leaks, nursery-school interviews and unemployability mean an Indian child’s most important decisions are choosing their parents and pin code wisely. The solution is not licensing but supply.”

FT: “How much value is AI really creating? Eye-opening changes to the speed and volume of work are not always translating into genuine productivity.”

Ron Friedman on his book Superteams: “I hope readers come away realizing that great teams are not the product of luck, chemistry, or hiring a few extraordinary people. In my research, I found that Superteams share three core strengths: (1) They get more done by managing their time, energy, and attention more effectively; (2) They make each other better, and (3) they keep improving over time. And the good news is everyone one of these strengths is learnable, which means by building the right habits, any team can dramatically improve its performance.”

WSJ: “For decades, institutions such as the U.S. Federal Reserve have made enormously consequential decisions—raising or lowering interest rates—based on incomplete and often delayed information. The Fed must adhere to mandates of full employment and price stability, but inflation readings arrive weeks after the fact. Employment statistics are revised months later. This leaves policymakers operating in a world of uncertainty, interpreting imperfect signals and using models undermined by gaps in real-time knowledge. The result is that central banks sometimes wait too long to raise or cut interest rates in the face of price changes in the economy. AI has the potential to change that.”

The Intent You Rent Back (Part 7)

The Intent Stack.

The three owners are not a menu to choose from; they are an order to follow. The cleanest way to hold that order is as a single structure — the Intent Stack — which ranks every source of intent a brand can act on, from the cheapest and most owned to the most expensive and most rented.

Most brands build it upside down

At the bottom sits historical intent: what the customer has already bought and done, held in your CRM. Above it, owned intent: what the customer does on your surfaces, generated by Atrium. Above that, predicted intent: what the BrandTwin expects next, supplied by Meridian. Above that, shared intent: what the cooperative network can route, through NeoNet. And only at the very top, rented intent: what adtech sells back, the last resort. The trouble is that most brands have built the stack upside down. They begin at the rented top — performance media as the default engine of repeat business — and then wonder why margins thin year after year. They are paying auction prices for a signal they could have owned, predicted or shared at a fraction of the cost.

Rebuild it in order

Rebuilding the stack in the right order is the whole of the doctrine in one move. Own what you can see, because engaged customers are generating intent on your surfaces now. Predict what you can model, because much re-entry is a knowable window, not a surprise. Share what the network can route, because your silent customers are alive on partner surfaces. Rent only what remains — the genuinely unreachable intent that nothing cheaper can see. Each rung you build downward from the top removes spend from the most expensive layer and moves it to a cheaper, identified one. The Intent Stack is not a new channel or a new product; it is a sequencing discipline that tells a CMO, for any given customer, which rung to reach for first. In practice a brand rarely builds all five rungs at once; it builds downward from wherever it starts today, reclaiming one layer of spend at a time.

What NeoNet really competes with

The stack also clarifies what each layer actually competes against — and the sharpest case is NeoNet. Its competitor is not another email ad network, not affiliate marketing, not display, not retargeting in the narrow sense. NeoNet competes with adtech’s single most valuable monopoly: cross-surface intent visibility. Google knows when people search; Meta knows when interest stirs; Amazon and the marketplaces know when comparison begins. The brand knows its customer’s history but not always the customer’s present. NeoNet connects history to present without handing the future to the auction — it is the one rung that matches adtech’s cross-surface advantage while keeping the economics on the brand’s side. That is why it sits directly below rented intent in the stack: it is the last owned-or-shared option before a brand is forced to pay the tax. Seen this way, NeoNet is less a media channel than a co-owned answer to the question adtech has monetised for twenty years.

Thinks 2009

Marina Nitze:”Crisis engineering is harnessing a crisis to make rapid transformational change. A complex system is any system that’s made of humans and computers. The computer part is generally pretty easy to change. And the human part is very difficult to change. But if certain crisis conditions are present, you can transform the human part of a system very rapidly. Crisis engineering is about how to recognize those indicators and then harness that moment to make rapid change against the human part of your system, and probably the computer part, too.”

FT: “Emboldened by big leaps in AI’s software programming capabilities over the past six months, tech industry leaders and researchers are increasingly confident that an AI that can improve itself with little to no human input is within their grasp. These self-taught AIs could keep on building new, more powerful versions of themselves. Successive generations could repeat the trick over and over again, adding new capabilities or making themselves run more efficiently. This flywheel effect, known in the industry as “recursive self-improvement” (RSI), could quickly lead AI beyond the large language models that underpin Google’s Gemini, OpenAI’s ChatGPT and Anthropic’s Claude, into uncharted territory. AI’s optimists believe it is the key to so-called superintelligence, or the point at which AI surpasses the abilities of the human mind.”

David Oks: “Why did China get rich, and India didn’t? What explains the Sino-Indian divergence?…China’s explosive growth wasn’t simply a matter of “freeing the markets,” reducing the role of the state, and announcing that it was now glorious to get rich; nor was it simply a matter of government intervention to support the manufacturing sector and subsidies for favored companies. China succeeded because it spent decades on the basics of human development and social modernization. India did not. The rest is just commentary.”

Ben Thompson: “How many companies could actually employ that cash in a way that generated a high rate of return? It’s hard to imagine a better option than Google. The company is not only investing in AI, but has optionality in terms of outcomes: its Services business benefits from the investment, it is in contention at the model layer with Gemini, and it can sell capacity to the frontier labs. Moreover, that capacity has a sustainable cost advantage because of TPUs, which means that in a world where compute becomes a commodity — as hard as that is to imagine right now — Google is the hyperscaler that is poised to make the most profit.”

The Intent You Rent Back (Part 6)

The three states of post-purchase intent.

There is no single solution to post-purchase intent because there is no single intent state. After a purchase, a known customer sits in one of three states, and each needs a different owner. Mistaking which state a customer is in is how brands reach for the wrong tool and end up at the auction.

Engaged — manufacture owned intent

If the customer still opens, clicks, taps, browses or replies, the brand does not need to detect intent from outside. It needs to create a surface where intent can keep appearing inside. That is the job of owned attention. A conventional promotional email waits until the brand wants to sell; a NeoMail earns attention before the brand needs to sell, giving the customer a reason to open — a useful idea, a quiz, a recommendation, a small reward, a moment of recognition. Over time the inbox becomes a living surface rather than a dumping ground for offers. And every meaningful action — a click, a tap, a saved item, a preference, a reply — becomes telemetry: identified, first-party signal about what the customer is thinking and may need next. Owned attention is the only owned intent signal, which is why attention sits upstream of transactions. This is Atrium’s first job: manufacture low-cost, continuous, identified attention so that intent never has to be rented later.

Predictable — anticipate before adtech detects

Some intent does not need detecting at all, because it can be anticipated. Replenishment cycles, renewals, seasonality, usage cadence, life events, festivals, travel, expiry dates and household rhythms all create predictable windows. A brand may not know the exact day a customer will search, but it can know the likely week or trigger. This is where intelligence earns its place. A BrandTwin, fed by the Customer, Product and Decision-Trace Context Graphs and the relevant world data, models the next likely intent window and acts before the customer ever appears on Google. The shift is from catching the signal to anticipating the window. A pet-food brand should not wait for dog food delivery to be typed into a search bar; an insurance brand should not wait for the renewal comparison to begin. If the timing is modellable, the brand should move first. This is Meridian’s job: convert predictable re-intent into lifetime value before that intent becomes expensive.

Silent — route through the network

The hardest case matters most. The customer is known but silent: no opens, no clicks, no visits, no replies. Owned attention has failed and prediction is too coarse — yet the customer is in-market somewhere else. This is the moment adtech was built to monetise. But there is another possibility: your silent customer is alive on someone else’s surface. They are reading a partner brand’s email, using a partner app, engaging with a non-competing brand in the same life moment. Their attention is not gone; it is simply not with you. NeoNet turns that fact into a recovery system. The old answer was an auction — upload the audience, bid for the user, pay the tax, hope the platform finds them. The NeoNet answer is routing — recognise that a known-but-silent customer has re-engaged on a trusted partner surface, and route a relevant recovery message through that surface before the brand falls through to paid retargeting. The network need not share raw data; it shares a decision — this person, silent for Brand A, is active in a trusted context where a relevant Brand A message can be shown. Recovery stops being a retargeting problem and becomes a routing problem. And routing should always be cheaper than bidding.