Search, Discovery, Pills, and Portals

Source: Venkatesh Rao, Contraptions · 2026-06-22

The distinction between search and discovery appears straightforward. Search connects people to things they already want. Discovery introduces them to things they did not know they wanted. This distinction underlies much contemporary thinking about marketing, recommendation systems, information architecture, and social media. Searchability is treated as a property of retrieval systems. Discoverability is treated as a property of feeds, recommendation engines, and social networks.

The distinction is useful, but incomplete.

Much of what is currently called discovery is not discovery in any strong sense. Recommendation systems rarely generate genuinely novel desires. More often, they accelerate the recognition of desires that are already latent. The user who encounters a recommendation for a restaurant, a book, a tool, or a short-form video often experiences the encounter not as surprise but as confirmation. The reaction is not “I did not know such a thing was possible,” but rather “that is exactly the sort of thing I was about to look for.” Discovery, in this sense, is anticipatory search. It surfaces tomorrow’s query today.

This suggests a first distinction. Search and discovery both operate within what might be called the future probable. They assume a relatively stable motive structure and work within it. Search satisfies existing motives explicitly. Discovery satisfies them implicitly. The difference is one of timing and awareness rather than substance.

Viewed dynamically, search is essentially non-perturbing. The user has already selected a destination. Search solves a routing problem. It reduces friction between desire and fulfillment. Discovery introduces a perturbation, but a damped one. It influences local path selection without substantially altering overall direction. A person who discovers a new snack food, podcast, or fashion trend may change behavior for a time, but the underlying motives remain unchanged. The perturbation remains contained within the same basin of attraction.

This perspective shifts attention away from information retrieval and toward the structure of adjacency. Why do certain things become visible to us rather than others? Contemporary recommendation systems rely heavily on mimetic adjacency. Things are nearby because people like us have encountered them. Collaborative filtering, social recommendation, and algorithmic feeds all operate according to this principle. The resulting discoveries are fundamentally self-referential. The organizing principle is derived from a model of the user.

Other environments rely on different forms of adjacency. Libraries offer an instructive example. The experience of wandering library stacks differs from browsing a bookstore, whether corporate or independent. A bookstore is organized around anticipated demand. Even the most curated bookstore remains oriented toward what somebody expects people to want. A library classification system is organized around an ontology. Books become adjacent because a bureaucratic scheme places them adjacent. The resulting serendipity is not random. It is structured by a classification system that is largely indifferent to the preferences of the visitor.

There is, however, another form of adjacency that is neither mimetic nor administrative. It is stigmergic. Things become adjacent because paths repeatedly intersect. The hot dog vendor happens to stand beside the falafel vendor. The coffee machine sits beside a hallway. A conference reception happens to place a historian beside a cryptographer. The resulting associations emerge through accumulated traces of movement rather than through either classification or preference. Stigmergic environments function as external associative memories. What becomes linked is determined by traffic patterns. Cities, campuses, conferences, and neighborhoods often derive much of their intellectual productivity from this mechanism.

At this point another distinction becomes necessary. Not all perturbations are equal. The magnitude of a perturbation is often a poor predictor of its long-term consequences. A large detour may produce no lasting effects. A tiny divergence may prove decisive. A driver who exits a highway to buy gasoline experiences a substantial local deviation while remaining on the same overall journey. A driver who chooses one of two nearly identical roads may inadvertently enter a new town, encounter a different environment, and ultimately abandon the original plan altogether.

The ε/δ perspective offers a useful way to think about this. As argued in the essay ε-δ Thinking, “The continuous, or ε/δ view of the world is fundamentally built around the fiction of becoming.” Small differences do not necessarily lead to small outcomes. Under certain conditions, “inputs that are too close to tell apart result in outputs that are radically far apart.” The important variable is therefore not perturbation magnitude but perturbation leverage. What matters is whether a perturbation occurs near a bifurcation structure.

Search and ordinary discovery mostly operate within stable regions of possibility space. Their ε/δ relationship is well-behaved. Small inputs produce small effects. More interesting phenomena occur near unstable equilibria, where tiny perturbations can produce large trajectory divergence.

One candidate for such a phenomenon is what internet culture calls a pill. Unlike discovery, a pill does not merely connect objects to motives. It alters the relationship among motives themselves. The common structure of ideological, religious, cultural, and lifestyle pills is not the creation of new desires but the reorganization of existing ones. A pill legitimates some motives while delegitimating others. It supplies permission structures, narratives, exemplars, and communities that allow a previously subordinated motive to become dominant.

The subjective experience is often one of recognition rather than transformation. Individuals rarely report acquiring entirely new desires. More often they describe the experience as discovering that desires they already possessed are legitimate. The underlying operation resembles a change in government more than the appearance of a new political party. Motives already present within the self acquire new authority.

This helps explain the durability of certain brands, movements, and identities. A consumer attached through utility can be displaced by a competitor offering slightly better utility. A consumer attached through preference can be displaced by the next cultural fashion. A consumer attached through identity is more resistant to churn. The relevant competition is no longer another product but another identity. Identity changes more slowly than preferences, and therefore supports more durable forms of loyalty.

Yet the more closely one examines pilling, the less radical it appears. A pill does not create a new self. It selects among existing possible selves. It provides legitimacy and social proof for an identity that was already latent. Its operation remains fundamentally one of selection rather than creation. It answers questions of being rather than becoming.

This realization points toward a final distinction. In Portals and Flags, a flag represents a move that stabilizes territory. A portal represents a move that enlarges territory. A portal offers “a more fertile way of thinking” that promises “an indefinitely extended stream of surprises within an ever-widening scope.” It does not recruit people into a worldview so much as create routes into new worlds. It can “turn it into a portal to a hidden universe of thought.”

Seen in this light, pilling is actually closer to flagging than to portalling. A pill stabilizes an identity. It strengthens a worldview. It recruits individuals into an existing regime of meaning. A portal does something different. It enlarges the space of traversable possibilities. Rather than asking which identity should dominate, it creates pathways among identities, disciplines, communities, or modes of thought.

Libraries often function this way. So do certain conferences, intellectual institutions, and historical projects. The original Whole Earth Catalog connected domains that ordinarily remained separate: ecology, engineering, computing, architecture, and self-sufficiency. Its value lay not in recruiting people into a single worldview but in creating routes among many. The same is true of environments rich in administrative and stigmergic adjacencies. Their purpose is not to stabilize identities but to create opportunities for movement.

The distinction is subtle but important. A flag answers the question, “Who am I?” A portal answers the question, “What worlds can I move among?” Flags stabilize. Portals enlarge. Flags recruit. Portals connect.

The original question about searchability and discoverability therefore turns out to have been too narrow.