The event is billed, with a kind of frictionless earnestness, as a “Founder Offsite,” though it is neither offsite (it takes place in a glass-walled co-working space whose chief architectural feature is its refusal to acknowledge walls) nor, strictly speaking, for founders (the ratio of actual incorporated entities to people wearing Allbirds and speaking in terms of “velocity” is approximately 1:14). Still, the name persists, which is itself instructive: in this ecosystem, naming is often the closest thing to doing.

There is free cold brew, which has the faint metallic tang of something dispensed through tubing optimized for scalability rather than flavor, and a buffet of small, aggressively photogenic pastries whose primary function seems to be appearing in Instagram stories captioned “building the future.” The attendees circulate in loose orbits, clutching laptops like flotation devices, and speak in a dialect that is both highly specialized and curiously contentless. One hears phrases like “leveraging synergies,” “AI-native workflows,” and “democratizing access,” all of which seem to gesture toward meaning without ever quite committing to it.

The ostensible purpose of the gathering is to workshop products that will, in the words of the event page, “reshape how humanity interacts with itself.” What this translates to, in practice, is a series of demos involving software that sits, barnacle-like, atop other software. There are tools to summarize emails that themselves summarize meetings that themselves exist largely to justify the existence of the tool. There are dashboards for tracking engagement with dashboards. There is, in one memorable case, a pitch for an AI that “helps you ideate faster,” which, upon further inquiry, appears to generate lists of ideas indistinguishable from those already being produced by the humans in the room, only faster and with slightly worse grammar.

It is tempting to dismiss all this as harmless—after all, no one is being forcibly subjected to a pitch deck, and the worst-case outcome seems to be yet another subscription one forgets to cancel. But there is a more uncomfortable question lurking beneath the surface, and it has to do with what, exactly, is being optimized here.

Because if you listen closely, beneath the buzzwords and the polite laughter at jokes about “pivoting,” there is a kind of quiet desperation. Many of the attendees have left stable jobs to be here. They have convinced themselves, and often others, that what they are doing matters—that building a slightly more efficient way to schedule meetings or a marginally more engaging way to scroll through short-form video is, in some attenuated sense, contributing to human flourishing. And perhaps it is, but only if one defines “flourishing” in a way that is both extremely narrow and highly contingent on continued venture funding.

One founder, who I will call K., explains his product to me with great enthusiasm. It is, he says, “a platform that uses machine learning to optimize personal productivity by analyzing your digital exhaust.” When I ask what this means in concrete terms, he describes a system that monitors your app usage, your keystrokes, and your calendar, and then suggests ways to be “more efficient.” I ask what happens if the most efficient use of one’s time is, say, to close the laptop and go for a walk. He laughs, a little too quickly, and says, “Yeah, but we’re not really targeting that use case.”

This is, I think, the crux of it. The system is not designed to question its own premises. It assumes that more engagement, more output, more optimization are inherently good, and then sets about maximizing them with impressive technical sophistication. The possibility that these metrics might be orthogonal—or even antagonistic—to well-being is treated as either a joke or an edge case.

There is an analogy here, though it is not a perfect one. In certain industrial processes, one can become so focused on increasing yield that one loses sight of what is being produced and why. The machinery becomes more efficient, the throughput increases, and yet the end product remains, in some fundamental sense, unchanged or even degraded. One ends up, in effect, with a highly optimized system for producing something that may not be worth producing.

None of this is to say that everyone in the room is insincere or deluded. There are, scattered among the pitch decks and the networking sessions, moments of genuine curiosity and even doubt. A developer confides that he is not sure his product is “actually useful.” A designer wonders aloud whether “we’re just making it easier for people to waste time more efficiently.” These moments tend to be brief and are often followed by a return to safer topics, like growth metrics or fundraising strategies.

By the end of the day, the cold brew has run out, and the pastries have been reduced to a few forlorn crumbs. The attendees disperse, some to after-parties, others back to their apartments where they will continue to build, iterate, and deploy. The language of the event—its talk of reshaping humanity—lingers, even as the actual artifacts produced seem unlikely to bear that weight.