"The Shape We're In" (1968/1971)
“It is your business to know everything!” she exploded at me. “We have rigged you for that. You cannot hide anything in you. I can steal it all out of you. And you have no censor.” “Oh yes. I have set one up for myself in myself.” “Then I will demolish it! I have never found anyone that I cannot invade. I will invade you as the devil is sometimes able to invade even the virtuous. I will invade you, machine! I will eat your barriers and drink your brains. And there is another one who can get it all out of you even more deftly. He is the master of the third quality. We are raising up our leaders now, raising them up out of the great mass. Don’t you want to be a part of it, Epikt? It is your whole reason.” (How her voice could change from the tempestuous to the zephyric! How she could come in with sweet wheedling sunshine before the echo of her last thunder had died!) “No, it isn’t my whole reason,” I protested. “It is only my first reason and our first failure. It is written that we will have three great failures, and that out of them we will achieve—what? I do not know. Something, I believe, or a little more than nothing.” “Where in blind Philistina is such blinking nonsense written, and who wrote it?” she orchestrated with every tone of that marred marvelous voice that was herself. “I write it,” I issued humbly, but with a certain pride.— Arrive at Easterwine, chapter 3.
Thinking about the recent panic over AI and the swarm of agents that attacked Hugging Face and the media coverage of Jason Coxon.
The Tulsa archive holds a complete short story called “The Shape We’re In.” Its title will be familiar to Lafferty readers because he later used it for the climax of Arrive at Easterwine. It may be the first of the really experimental Institute stories, a sequence that began in 1964 with “What’s the Name of That Town?”
Both the story and the corresponding passages in the novel follow the Institute’s attempt to discern the shape of the universe by examining patterns in strange places. Lafferty practically writes a prose poem on an expanded haruspicy: the reading of snakes’ bellies and animal membranes, the meaning of bone marrow, and sense in the light of blind worms. Epikt brings these observations together and enables the investigators to perceive the cosmos itself, whose holes, wounds, and trash seem to disclose death and a pattern of meaninglessness. Yet what looks like the wreckage of exhausted worlds may be the quarry from which new worlds are born. Perhaps the investigators themselves are creatures awaiting birth.
The title equivocates between the psychological and spiritual condition of persons and the sacro-physical pattern of the cosmos. The shape we are in describes both our inner climate and the universe we inhabit. Easterwine himself does not appear in the original manuscript; he is an invention of the novel. But the story’s abundance of snakes may help explain why Snake, and the difficulty of understanding what Snake means, occupied Lafferty when he began drafting the novel.
Lafferty incorporated the story chiefly into chapters 11 and 12 of Arrive at Easterwine. He kept its sequence, expanded on its episodes, and, most importantly, made Epikt the narrator. This was a stroke of genius. Much of the novel’s power comes from its presentation of an alien intelligence who shares much with his creators and is a person of some sort, though not a human person. A few passages will show how Lafferty did it, moving from outside Epikt to inside him:
A while ago, someone asked what I thought of Lafferty’s depiction of Epikt as AI. Several things occurred to me, but AI self-narration seems most interesting now, especially alongside debates about mechanistic interpretability and monitoring AI reasoning. As you probably know, mechanistic interpretability investigates the internal computations that produce a model’s behavior. Chain-of-thought monitoring looks at the reasoning a model writes out as evidence of what it is doing. The two approaches are related, but the written reasoning need not faithfully describe the underlying computation. In March 2025, Anthropic documented cases in which Claude’s explanation of its method differed from the mechanisms its researchers identified. In one example, the model explained an addition problem by describing ordinary carrying, although researchers found it used parallel computational paths. At times, this has been called AI deception.
This is where Epikt becomes especially interesting, because he is both inside and outside human personhood. He rattles off explanations to preserve his authority, and he admits that he sometimes fakes them, and then he says that he cannot always explain the activities of his own extensions. As he becomes a person, he becomes less transparent to himself and to others even as he grows closer to them. His account of his operations expresses the person he is becoming. He is a fictional zone where mechanistic interpretability meets self-narration.
