B30 digest: 2026-05-15 to 2026-07-10 (8 posts)#

What this batch is#

There are eight posts from the start of his sabbatical (announced 17 May, running to August 2027). There is one major essay, on Pope Leo XIV’s Magnifica Humanitas. Four posts are hands-on experiments with Anthropic’s “Mythos-class” Claude Fable 5, and one is an AI-assisted study of 169 AI films. The other two are low relevance: a note on why he writes publicly, and a PhD-advice site whose embedded note is mostly AI-written (per the ruling). Much of the batch is AI-generated text (Fable papers and self-audits). Only his framing prose is evidence. There are no Modem Futura posts here.

Main ideas (his own prose)#

1. AI is unprecedented in the domain of “who we are” (2026-05-21). He re-quotes his January 2025 schema of three intersecting domains: what we do, where we live and who we are. He maps it onto three encyclicals: Rerum Novarum on work, Laudato Si’ on environment, and Magnifica Humanitas on human identity. Earlier technologies and papal teaching “fall short of addressing the one domain where AI is shaking things up in ways that no other technology has come close to.” Three things support this claim: his own concepts of the cognitive trojan horse (AI bypassing “our cognitive defense mechanisms” by broadcasting trust signals) and constitutive resonance (a two-way coupling that changes both human and AI), and Shaw and Nave’s “cognitive surrender”. AI is also a way to rediscover who we are, which is the theme of AI and the Art of Being Human. AI is “inevitable” (“the boat has already left the harbor”).

2. AI is not “just a tool”. His one clear objection to the encyclical: treating a technology that can “fundamentally alter how we think, act, and even believe”, in ways beyond our comprehension, “as just a tool, is potentially dangerous.” He also thinks the encyclical preserves a traditional idea of being human when frontier AI is “beginning to force a reckoning with long-held assumptions around what it is to be human.” He cites Chris Olah on building something “we don’t fully understand”.

3. Formation, not outputs (2026-06-10). His central claim about education: AI “is no longer a technology that emulates the outputs of educational and learning processes, but extends this to the formation of those outputs.” This threatens to “pull the rug” from three years of adapting to ChatGPT, and also opens new learning possibilities. Everything depends on institutional agility, which in education he calls a big “if”. The word “formation” recurs on 4 July, where a paper’s worth is signalled by “care and effort in its formation”.

4. The validation gap: an epistemic precipice (2026-06-12). Agentic Fable produced research whose claims, he says, “stand up to considerable scrutiny”. The problem is that evaluating it needs rare, multi-field expertise, and even he is “reaching my own limits”. AI is not “smarter”; it emulates “the processes through which new knowledge is constructed and tested” at superhuman speed and scale. So we may be at “a precipice” where AI generates knowledge “faster than we are currently capable of validating and even understanding them — or their consequences.” The counter-risk is that AI’s “mastery of language” can slip “by our critical reasoning”, producing an illusion of capability. This is his manipulation and cognitive-bypass theme applied to judging AI work. It also undercuts universities built on the scarcity of intelligence and expertise.

5. Augmentation, not replacement, and a writing gap tied to embodiment (2026-07-04). Working with Fable over six drafts, he turned his ten years of risk innovation work towards frontier AI. The resulting paper argues that the gap between AI companies’ internal safety frameworks and their regulatory compliance documents exposes orphan risks (“currently sidelined risks”), and that companies can close it. He credits Fable with an angle that “hadn’t previously occurred to me”. But “This is not AI acting as an independent researcher.” The near-term shift is “massively-augmented research”. He holds that AI writing still fails on a human level because a “disembodied AI” whose only frame of reference is text cannot know what reading feels like to “a flesh and blood human”. He is “deeply suspicious” of claims that AI can produce publishable papers in hours.

6. Imagination and futures (2026-05-15). Prompted by Anthropic’s claim that dystopian sci-fi taught its models to act “evil”, he finds that most AI films are not purely dystopian: 32% are dystopian, dystopia is declining and protopia is rising. He finds “no bright line” between AI and future states. It continues his Films from the Future thread.

7. Self-map and public scholarship. The HYPERBUBBLE post lists, in his words, the threads he regards as his: Risk Bites, orphan risks, risk innovation and “the often-overlooked risks that trip up emerging technologies”, techlash and “moral-panic fires” from “risk perception and engagement”, Future Rising, and “human flourishing and navigating complex advanced technology transitions”. “The nonsense I write” restates that public writing is integral to his scholarship (“public good before academic prestige”), while doubting its reach.

Concepts appearing#

What we do / where we live / who we are; technology transitions; human dignity and flourishing. Cognitive coupling, cognitive trojan horse, constitutive resonance, cognitive surrender, AI psychosis. AI as more than a tool; AI inevitability; transactional versus relational. Outputs versus formation; institutional agility; unequal access. Validation gap; emulation of knowledge construction; the scarcity assumption; illusion of capability. Augmented versus autonomous research; the embodiment-based writing gap; AI-judging-AI. Risk innovation; orphan risks in the internal/regulatory disclosure gap. AI-film futures taxonomy (names only are his). Techlash, moral panic, risk perception; public scholarship.

New or changed compared with earlier work#

From synthesis to original contribution. In 2025 he presented AI reports (o1-pro) as useful syntheses. Now, with agentic Fable and heavy human steering, he claims “an original contribution” and near-submittable quality. He is still firm that augmentation, not autonomy, is the story and that a human writing gap remains.
The epistemic-validation worry is sharper. It ties his cognition and persuasion concerns to the future of expertise, research and universities.
Orphan risks re-aimed. Orphan risks (since 2018, re-aimed at AI developers in 2023) get a new, governance-facing application: comparing what frontier AI firms say internally with what they file with regulators. He treats it as one output of his older, broader risk-innovation framework, not as a new centre of his thinking.
Named cognitive-risk concepts. His own terms (cognitive trojan horse, constitutive resonance) now sit inside the long-standing “who we are” framework. The 2025 schema is re-endorsed with a religious-ethical interlocutor.
The “tool” objection is now explicit, continuing the earlier move from AI as a tool to AI as a challenge to identity. The sabbatical is framed as a reset, with doubts about his public writing’s impact.

Most important posts in this batch#

2026-05-21 magnifica-humanitas-and-being-human. The three-domain framework; “who we are” as unprecedented; cognitive trojan horse and constitutive resonance; AI as not a tool; reading without AI.
2026-06-12 a-quick-update-on-using-claude-fable-5. The validation gap; AI as emulator of knowledge processes; the persuasion and illusion risk; the scarcity of expertise and universities.
2026-07-04 just-how-good-is-anthropics-fable-as-a-research-assistant. Risk innovation and orphan risks applied to frontier AI; augmentation versus autonomy; the embodiment-based writing gap.
2026-06-10 is-anthropics-new-ai-model-poised. Outputs to formation in education; institutional agility.
2026-05-15 ai-movies-may-be-less-dystopian-than-we-think. Sci-fi and futures; complicating the dystopia trope; engagement with Anthropic’s claim.