Beyond Staying Afloat

The question behind every product launch

The distinction that matters most in technology is not between products that work and products that do not. It is between products that people believe are important and products that people actually use. These are not the same category, and the gap between them is wider, more systematic, and more consequential than any serious person in the industry admits out loud.

There is a startup that receives extraordinary press coverage, raises significant capital, and earns thousands of upvotes from people who genuinely believe it is important. And then, without drama or announcement, the vast majority of people who downloaded it stop using it — not because it stopped working, but because the urgency that the launch generated was not the same thing as the need the product addressed. There is another product that almost no one covers anymore. Every morning, without being asked, hundreds of millions of people open it before they have eaten breakfast. They open it because not opening it would be a problem in their day.

The product that wins makes the cost of not using it genuinely higher than the cost of using it, and does this for reasons that are not going away.

What winning actually requires

A product wins when three qualities converge simultaneously. The industry treats them as separate dimensions and congratulates itself when a product achieves any one of them.

The innovation signal

The first quality is perceived innovation — the genuine excitement that a product generates when people first encounter it. It drives the social urgency to share, to write about, to upvote. It is also, in isolation, almost perfectly uncorrelated with the product's actual importance.

The utilisation signal

The second quality is actual utilisation — the state in which a product has become not just used but depended upon. This is the signal that is hardest to fake. A product can generate extraordinary enthusiasm from people who cover technology for a living and barely register in the daily behavior of people who use technology to live.

The longevity signal

The third quality becomes perfectly visible ten years later: durable longevity, the state in which a product remains genuinely relevant because the human need it serves has not changed, and every subsequent attempt to serve that need better has returned to the same product as the best available answer. The products that achieved all three defined a category that had never existed, served a need that had never been adequately served, and made the previous answer feel retroactively inadequate. Everything else is optimization.

The hype cycle: what the curve has been telling the industry for thirty years

Gartner published the first version of the Hype Cycle in 1995. It remains the most accurate description of how technology moves through the world, and one of the most consistently misread, because people read it as a warning about the trough and miss what it is saying about the plateau.

A technology emerges. Attention and capital follow. The narrative exceeds the reality. Most products do not survive the gap between what the narrative promised and what the usage data reveals. The ones that do climb toward the plateau of productivity, where they become genuinely useful and genuinely important without generating the attention they once did. The largest technology companies have been at their respective plateaus since approximately 2010. The plateau is where staying afloat looks like success. It is also where evolution stops.

What the cycle cannot chart is the new innovation trigger above the current plateau. The companies sitting on the plateau are not the ones who will produce it. They are built to maintain the plateau, and they are extraordinarily good at that.

Twenty-five years of the same ideas

Take the largest technology companies in the world. Describe in one sentence what they do. Then describe in one sentence what they did when they were founded. The distance between those two descriptions, in fundamental purpose, is very small.

A search company launched as a better way to find information through a better algorithm and remains exactly that, incomparably faster. A social network launched as a way to connect with people you know and see what they share, and two decades later that is precisely what it is. An online store launched as a way to buy things and have them delivered, and at incomprehensible scale, that is what it remains. A professional network launched as a resume connected to other resumes, organized around jobs, and that is precisely what it remains. These companies did not change what they are fundamentally for. They changed how well they do what they were always for. The gadgets evolved. The purposes did not.

The three mechanisms that make the plateau self-sustaining

Acquiring the evolutionary threat before it can threaten

Every major platform sitting on its plateau has done so in part by purchasing the companies that were building the most serious alternatives. Each acquisition was rational. The aggregate effect is that every genuinely evolutionary product that achieved enough traction to be threatening was acquired before it was large enough to complete the evolution it was beginning. What could have been a new curve became a feature announcement.

Feature velocity as the performance of innovation

When a company cannot evolve its fundamental category, it releases features. Adding an AI assistant to a search engine is a feature. Adding short-form video to a feed is a borrowed feature. Each is a genuine improvement. None changes what the underlying product is for. This is institutional self-preservation through activity that is productive enough to justify the spend and novel enough to justify the coverage, but not disruptive enough to threaten the revenue.

The network effect as civilizational lock-in

A social platform with several billion users does not hold them through product quality. It holds them through the accumulated archive of every relationship, memory, and piece of social history that the product made possible over two decades. The switching cost is not learning a new interface. It is the loss of the social infrastructure of a connected life. The evolutionary threat will not come from a better version of what they built, but from something that makes the question those platforms answer feel less central to how people want to live.

Revolution and evolution: the category error that cost the industry a decade

A revolutionary technology makes something that already exists significantly better — faster, cheaper, more accurate. An evolutionary technology makes the previous answer to a human need obsolete. The automobile did not make horses faster. The photograph did not help portrait painters improve. The internet did not improve newspaper printing. These transitions did not win on the terms of what came before. They redefined the terms.

The smartphone was the last genuinely evolutionary technology at consumer scale. Everything significant introduced since is revolutionary rather than evolutionary. Generative artificial intelligence is the most sophisticated revolution in text production in the history of human communication. It makes producing text cheaper and faster. The question — what do human beings want to say, to know, to create, to understand — remains unchanged. The technology that makes the previous answer to that question feel retroactively inadequate has not arrived yet. It is the thing worth building.

The conditioning: why brilliant people consistently miss what is directly in front of them

The professional culture of technology is organized around comparison. A new product is evaluated against the closest existing alternative. This framework is well suited to identifying revolutionary products and precisely wrong for evolutionary ones, which have no closest alternative and whose differences read as deficiencies rather than as signs that the comparison itself is the error.

In January 2007 the professional technology community evaluated the first iPhone against existing phones and found it inferior on the criteria that comparison generated. The analysis was wrong not because its data was incorrect but because its frame was. The question it answered was not "how do you make phone calls better?" but "what should a device that fits in your pocket and travels with you everywhere actually be able to do?" The products that matter most over the longest periods are precisely the ones that are hardest to evaluate using the tools the industry has built to evaluate products.

What should exist by now

In education, the transmission of knowledge no longer requires proximity, the pace no longer needs to be synchronized with thirty other people, and capability could be assessed continuously in authentic contexts. The result in practice is that the textbook is now a screen and the assessment still measures whether information was retained.

In healthcare, the conditions that consume the largest share of resources are visible in behavioral and physiological data years before they produce symptoms. The result in practice is that it is easier to book appointments and receive video consultations for conditions that would once have generated in-person ones. The paradigm — a condition develops, a person seeks care — has not changed.

In work, the eight-hour day was designed for factory tasks where output was proportional to time. The science of how people actually do intellectual work has been accumulating for decades. The result in practice is that some people now do the same work at home on the same schedule with the same expectations.

In human connection, the platforms built to connect people made weak-tie connection easier and have been associated with rising isolation and loneliness, because weak-tie connection was not the need. This is the most important unsolved problem in the technology industry.

What we are building: a community for the next phase of human evolution

The evolutionary step will not come from the institutions that are most successful within the existing paradigm. It requires a different kind of institution: one organized not around optimizing what exists but around investigating what should replace it.

The closest existing models are the intentional communities that have built functioning alternatives to the assumptions they questioned. Findhorn, founded in Scotland in 1962, has sustained several hundred residents through more than six decades and built twenty community enterprises from first principles. Damanhur, founded in northern Italy in 1975, has housed roughly a thousand people for nearly half a century in a complete civic infrastructure with its own governance, economy, education, culture, and monuments. What distinguishes these communities from every coworking space and accelerator that has claimed similar outcomes is that the people in them are not working in proximity. They are living in proximity.

The community being built — abroad, in a location selected for its ecological clarity, its altitude, and its distance from the cultural centers where the industry's assumptions are most densely concentrated — is organized around radical individualism within genuine community. Every person who joins is there to develop their own specific purpose and contribution, and is in the community because genuine community makes individual development faster, deeper, and more durable. The community does not define the individual's purpose. It creates the conditions in which the individual can find it.

The contemplative traditions that ground this work are not decoration and not wellness branding. Vipassana meditation, a twenty-five-hundred-year-old practice for the direct observation of the nature of mind, asks its practitioners to observe what is actually happening rather than accept a description of what should be. A platform for the intelligence of human sleep designed by people who have genuinely investigated their own consciousness in ten days of silence is more precisely conceived than one designed from the research literature alone. The practices and the platforms are the same project, expressed at different scales.

The long-term architecture is specific: energy generated on the land, food produced through regenerative agriculture, governance through participatory processes, economics through the platform revenues of the ecosystem and the direct activity of the community itself. This is not a vision of isolation. It is a vision of freedom.

One hundred platforms for what comes after the plateau

The directory that accompanies this letter presents one hundred startup concepts for the human experiences and civilizational domains that the technology industry has not yet seriously attempted to serve. Each addresses a domain that is simultaneously among the largest economic markets in the world and the most comprehensively ignored by the startup ecosystem. Each has the defining characteristic of the best startup ideas: reading about it, the only reasonable response is to wonder why no one has built it yet. The industry has been so successful at optimizing what it already occupies that it has not developed the habit of noticing what it has never entered. These are the things it has never entered. The work of building them begins now.

Warm regards,
Harshal Agrawal