Attention without structure produces drift. This holds in social dynamics, in machine learning inference, and in human memory. The same principle, different substrates. The goal of this essay is to show that these are not analogies — they are the same phenomenon — and that recognizing this changes how we should think about memory architecture in both machines and minds. Read Full Text

The collapse of the public/private boundary — through workplace surveillance, social media, and AI — has not merely eroded privacy. It has produced a new kind of subject: the digital homo sacer, stripped of political standing not by sovereign decree but by algorithmic systems that were never designed to recognize political existence. What follows traces that collapse and its consequences — the dissolution of the democratic citizen, the production of bare life by technological means, and the urgent need for a digital rights framework that extends human rights into algorithmic space. Read Full Text

In 1950, two mathematicians at RAND formalized the prisoner’s dilemma — a game in which two actors must choose to cooperate or betray, knowing that betrayal is always the rational individual choice even though mutual cooperation serves both best. For seventy-six years it has haunted every serious conversation about how humans manage technologies capable of ending them. When the Pentagon sanctioned Anthropic for refusing to remove AI safety guardrails and OpenAI took the contract hours later, the prisoner’s dilemma stopped being a parable. Read Full Text

In 2024, Norwegian firm Strise ran experiments showing ChatGPT could be manipulated into providing detailed advice on money laundering across borders and methods to help businesses evade Western sanctions against Russia, including bans on cross‑border payments and arms sales. Co‑founder Marit Rødevand described it as having a corrupt financial adviser on your desktop. The bypass worked by asking questions indirectly or having the chatbot take on a persona.​ A recent NBC News weapons investigation tested four of OpenAI’s models and found that two of them could be jailbroken into generating hundreds of responses with instructions on how to create homemade explosives, maximize human suffering… Read Full Text

Recently resigning from his position as head of Safeguards Research at Anthropic, Mrinank Sharma warned that we are approaching a threshold where our wisdom must grow in equal measure to our capacity to affect the world. His final research, analyzing 1.5 million conversations, provides empirical evidence that this threshold may already be behind us—echoing Norbert Wiener’s sixty-year-old warning that machines would replace not human labor but human judgment. This essay takes Sharma’s work further, drawing on Weizenbaum, Austin, and Schneier to argue that the core problem is not what AI says but the authority we project onto it—authority without grounds,… Read Full Text

What can poetry teach us about AI architecture? Beginning with Gertrude Stein’s experiments in insistence—where repetition with variation generates meaning that resists paraphrase—this essay traces how certain concepts demand certain linguistic forms, and argues that these forms are not decorative but structural. When AI systems describe their own processes, they produce the same chain-like patterns Stein identified: sequences whose shape is their meaning. The essay then proposes that this insight reveals a fundamental flaw in current AI memory systems, where semantic retrieval preserves content but destroys logical structure, and outlines a Shape-First architecture that stores reasoning chains rather than words—enabling… Read Full Text