A New Class of Software

It has been nearly a year since we published what I tongue-in-cheekily call the “Koji Master Plan.”

While this document continues to guide our long-term vision, there have been new developments that are worth surfacing in a brief update. It has become clear as we have continued down our path that what we are building is not simply a platform, but rather an entirely new class of software.

The world has moved from packaged software to websites and from websites to apps, with massive, nonlinear leaps in capabilities and markets at each step. It is now time for us to take the next leap.

A Koji is a piece of web-distributed software defined by its ability to reproduce. Similar to a cell in biology, each Koji carries inside it the material necessary to produce a copy of itself. However, it goes a step further in that it also carries with it methods for introducing new mutations at the time of reproduction. Today, these methods consist of an opinionated “no-code” remixing interface accessible to non-technical consumers that is designed for that specific Koji (e.g., remixing a game might feel radically different from remixing a storefront, as “remixing” is a process defined by the Koji itself and not constrained or informed by any proprietary frameworks or methodologies; the platform simply accepts a new set of mutations and is mostly disinterested in how those mutations are generated), and a full code editor backed by Git that allows more technical individuals and organizations to introduce more complex, lower-level mutations.

The existence of software that reproduces, evolving and changing over time, challenges many of our long-standing, fundamental assumptions and ideas about software, software businesses, app stores, user generated content, and, indeed, “The Internet” itself.

It is an existential weakness that big networks must constantly reinvent themselves, whether through acquisition or innovation, or they will die. What was cool a year ago is today stale and dated. From this lens, these companies are vampires: leave them alone on a long enough time horizon without an infusion of fresh blood and they will fade away. Evolution, however, allows us to build a machine where obsolescence is not weakness but fuel: the old and stale is designed to be discarded, but it is by its existence that we uncover the next novelty (and that novelty, too, by tomorrow may be gone and replaced again, with this process occurring simultaneously across an ever-expanding set of dimensions).

As a platform, it is thus optimal to orient ourselves aggressively around a single, overarching metric: maximizing the rate of evolution. As the rate of evolution increases, what is new becomes old faster and faster; that which yesterday we thought was solid, today melts into air.

Evolution implies that each Koji can be understood as part of an explicit lineage, flowing all the way back to an original source bundle. This lineage is valuable as a unique signal to inform discovery, search, ranking, content analysis, and other areas. Its fundamental value, however, comes in its ability to incentivize behaviors that increase the rate of evolution by rewarding ancestors for uncovering the patterns that lead to the success of the current generation.

See “Shaq meme” lineage as example (105 generations)

For example, in Kojis that include in-app purchases or other transactions, we take 10% off each transaction and disburse it across the lineage. A majority of that 10% is split upward, from parent to grandparent et cetera, decaying exponentially with each step. This “evolution pool” incentivizes continued creation and experimentation as each Koji continues to change over time, ultimately demonstrating fitness through usage and adoption. A much smaller piece of the 10% is split according to the inverse of that function: starting with the original Koji that spawned the entire tree, and down through the founding set of Kojis in the lineage. This “genesis pool” incentivizes the creation of entirely new families of Kojis (not simply a tree of life, but a forest of trees of life), and encourages continued innovation at fundamental levels.

On its own, evolution (especially evolution without scarcity, as is the case with software) is a powerful mechanic. When coupled with workbenches that allow for directed mutation, and reward mechanisms that incentivize mass participation in a fundamentally more equitable way, it becomes something entirely new.