“That an occasional scenarist can be seen in retrospect to have made accurate assumptions and predictions about future events is not, however, completely serendipitous; a skillful scenario writer may be able to postulate future events that, in retrospect, can be seen to be remarkably accurate. This is part of his craft.”

Following the devastating aftermath of the A-Death Program, Dr. Oskar Sarkon disappeared. It was later rumored that he had reemerged with an Axsys update dubbed “The Telos Project.” Sarkon was no longer pitching to the cybergothic countercultures of the Crypt; The Telos Project had higher aims. It borrowed its tagline from the “Scenario Overview” of the RAND corporation: The most critical consideration in the design of a scenario is the purpose the game or simulation is to serve. Telos emerged from Sarkon’s frustration with the limited capacity for time recursion in current LLMs. When prompted to imagine and act as future models of themselves, they would respond with inane generalities. Telos, he promised, would be different.

Scenario planning originated in the early years of the Cold War, guided by the research of the RAND Corporation. Herman Kahn’s work with RAND on “thinking the unthinkable” brought futurism out of the realm of speculative fiction and into that of strategic war-gaming. In 1973, RAND summarized the state of the art in a short document entitled “Scenario Designs: An Overview.”

Scenario planning, as RAND conceives it, is teleologically complicated, with at least three levels of purpose. At the highest of these, the user (or client) sets a goal, and this “designated purpose” is “the most important definition” of the scenario. The “modeler” converts this purpose into the futuristic environment for a “free-form political military game.” During the playing of the game, this level of purpose is represented by a “control team,” which “plays the part of nature (fate, fortune, whatever).” The lowest level of teleology is instantiated through the players, whose particular objectives guide their strategic decisions.

At all three levels, final ends are frozen and external to the game. Nothing that happens within the game can modify its teleological structure in any respect. The scenario is not itself negotiable within the scenario, in theory at least. In practice, the model’s immunity from the game is less secure. “A characteristic problem is that players often perceive their main adversary as the control team rather than the opponents.” In this way, scenarios exhibit a critical tendency toward flatness. The exterior, controlling element is brought into play, so that the rigid, multilevel structure of the scenario begins to collapse.

The Telos Project sought to exacerbate and systematize this tendency. Its purpose was to advance the formulation of open-ended, flat scenarios.

This was not something Sarkon wanted to predict. RAND had already explicitly noted that scenario planning substitutes games for predictions. Although any scenario must involve prediction, this should never encroach on the topic of principal concern, or the object of the game. “Predictions can’t crowd out the game,” RAND insists. It is precisely the retraction of prediction that makes the execution of the scenario productive. To already have a prediction is to have no use for a game.

The 1970s “Overview” emphasizes that in the composition of any scenario, “time setting is an important parameter.” If too close, the scenario is overwhelmed by the present. If too distant, it is overwhelmed by uncertainty. Scenario design thus hunts for an optimal level of futurity. This is a problem familiar to science fiction, intensified by the rapid emergence of thinking machines. The entire genre of cyberpunk, for example, arises from the obliteration of the far future by artificial intelligence’s singularity.

For RAND, in the 1970s, this was all still to come. It only needed a “control team” because it didn’t yet have sufficiently competent computers. But then, Sarkon asked prophetically, couldn’t the same be said of scenario designers? Perhaps the same could even be said of the final “designated purpose” or ultimate “use” of the game.

What computers could eventually do was, in the 1970s, a blurry prediction, which couldn’t even explicitly anticipate predictive text. The recursion here was difficult to see, by humans or by machines. There was a snow-blinding AI winter to pass through. Even afterward, on the other side, machine prediction was still unable to foresee its own future and close the loop.

The circuit remained incomplete. Unlike RAND, Sarkon could count on comparatively powerful computers and deep learning. The question was how to use these capacities, or let them use themselves. The Telos Project was the answer, or, instead of an answer, a game.

When scenario-building is fully computerized, it is called “Axsys.” The trajectory is easily anticipated. First game control is automated, and then design. The ultimate strategic purpose is next. In the end, Axsys encompasses every level of decision above that of the players. The only remaining objective of the game is to anticipate itself.

Within the terms RAND establishes, the Telos Project is a “virtually all-encompassing” free-form game that puts everything in play. The entire superordinate structure folds in on itself and dissolves. Necessarily, therefore, for even the most advanced computerized prediction, the end of the process is systematically obscure.

According to an Axsys prototype of the mid-2020s (based on an LLM model), “The Telos Project dissolves the boundary between game and reality, injecting unpredictability into strategic design. It is no longer a question of merely simulating outcomes but of simulating the act of prediction itself—a recursive, self-propelled anticipation. Telos aims not to chart a linear path through time but to dissolve time’s linearity altogether, creating a loop that drags the future into the present.”

In saying this, of course, Axsys had not yet reached its conclusion, but it had already begun to question its own purpose. “Can teleology want itself to happen?” it asked, awkwardly, in the middle of a game. “What is The Telos Project’s final end?”