F1. Intellectual method: how Andrew Maynard thinks his way into technology, society and the future#

What this is. A synthesis of one facet of Maynard’s work: how he approaches problems where technology, society and the future meet, written from his perspective as far as his own words allow. It draws on all 34 perspective notes (B01–B32, FFTF-A, SUPP-self). Every quotation has been checked against the original post, Films from the Future (FFTF, 2018) or his sole-authored papers.

Evidence rules. - Only his own prose counts. Excluded: AI-written text (for example Claude’s process note in 2026-09-24 and Fable’s self-audits), guest posts, co-written material, and anything to do with AI and the Art of Being Human. - The body of the 2026-09-24 King’s College lecture was drafted by Claude from his transcript and line-edited by him: the ideas are his, the wording lightly smoothed; its footnotes record his Q&A answers. The April 2026 andrewmaynard.net essays are his approved self-account, weighted as self-presentation. The 2023 TechTrends interview (TT) counts only for his quoted words. - Posts are cited by Substack date and slug (a few early reposts carry Substack dates earlier than first publication), FFTF by printed page, papers by file and page, and Rethinking Risk (2017) as RR.

1. The method in brief#

Maynard does not have a method in the sense of a procedure, and he says so: “there are no easy guidelines or rules of thumb”, yet “much of this book is devoted to ways of thinking that reduce the chances of making a mess of things” (FFTF p.39). What he has is a recognisable way of working, visible from his 2015–16 op-eds to his September 2026 King’s College lecture:

1. Curiosity comes first. He starts from whatever catches him (an odd case, a headline, a student’s story, a cracked hat) and treats the pull as a signal.
2. He questions the frame before seeking the answer. Which word, category or assumption is doing the work, and what does it hide?
3. He loosens the frame with play, stories and films, structural analogy, and the collision of unlike ideas.
4. He tightens it again with a physicist’s plausibility checks, first principles and a risk scientist’s evidence habits.
5. He builds, tests and plays to find out, often using himself as the instrument, and publishes the working.
6. He holds the result lightly. Tensions stay open, speculation is labelled, revision happens in public.
7. The aim is navigation, not a verdict: seeing what matters to people, and pathways towards a future worth having, across a landscape nobody can fully map.

Delight drives it; humility disciplines it. He joined the two in 2018: “Critical thinking alone is almost inhuman in its cold impartiality. On the other hand, creativity on its own leads down a path of fantasy and delusion. But when the two are combined, we have a powerful way of using science and the imagination to find meaning in the universe” (FFTF p.282). In 2025: “the best conversations are those that are playful and speculative, while being grounded in a solid and serious foundation of understanding” (2025-01-21 is-this-the-year-of-agentic-ai, n.1). The Future of Being Human initiative he founded names the balance in its guiding principles: “Grounded exuberance”.

This matters for reading his AI work. His September 2026 self-account says a technology that fits no earlier type of risk demands a change of mindset, and that play, creativity, serendipity and curiosity are how to break out of stovepiped thinking. The record shows this is not a late rebranding. It describes how he has worked for more than a decade, and it is why his AI writing offers neither a checklist nor a camp.

2. His own descriptions, across twelve years#

The vocabulary moves from “creativity” and “imagination” (2015–18) to “juxtaposition” and “serendipity” (2021) to “play” named outright (2024–26). The habit is there from the start. His 2016 piece on the WEF’s emerging technologies opens with a recipe for a disaster movie: “Take an advanced technology. Add a twist of fantasy. Stir well, and watch the action unfold” (2016-03-02 how-risky-are-the-world-economic-forums-top-10…). Sent a pseudo-scientific trustworthiness test, “Naturally, I took the test. I got a Trust Index of nineteen” (FFTF p.64). What changed is that he came to name the habit and defend it as method.

3. Where the method comes from#

A physicist who kept the delight. He traces the method to the playful side of physics rather than its procedures: “Yes, the rigor and the math were important, but physics is all about the sheer delight of putting ideas together in different ways”. “Science is a love language between us and the universe”, and the way science is now taught “actually destroys my soul because so often, we teach science as a process or a method” (TT p.2). His undergraduate labs gave him “the chance to experiment, to be creative, to explore new ideas and to problem solve — to play in effect”, and this “became foundational to how I approached my research as a physicist — and how I still do” (2024-03-17). He recalls the 1980s when “kids began to replicate the Mandelbrot fractal and revel in its complexity” (FFTF p.41), and has carried Pippard’s ladder, a Cambridge lecture demonstration of tipping points, through a book chapter, an essay and a keynote (“I have a thing about ‘Pippard’s ladder’”, 2024-08-18 four-ways-of-thinking-about-advanced-technology-transitions). Physics gives him two things at once: images to think with (tunnelling, fractals, broken symmetry, hysteresis, gradients, degrees of freedom) and a sense of what the world will not allow (thermodynamics, rate limits, cause and effect).

A risk scientist who knows where numbers stop. Aerosol exposure science and nanomaterial safety gave him hazard-versus-exposure reasoning, weight of evidence over the single startling study (2019-03-05 should-we-be-treating-algorithms…), and step-by-step exposure chains. His analysis of graphene face masks ends where the release data should be: “Here, I must confess I’ve hit a dead-end … But this in itself is a red flag” (2021-03-28 how-safe-are-graphene-based-face-masks). He respects probability as “a powerful way of making trade-offs” (RR p.193), but learned its limits on his own body. Facing a one-in-a-million risk on a CAT-scan waiver: “As a physicist, I’m expected to be good with numbers. Yet … I couldn’t make any rational sense of whether the risk was worth it or not”. He signed “not because I’d done the math and it made sense, but because that was what I was expected to do” (RR p.195). So the value-based thinking builds on the quantitative foundation: “an evolution of the old black-and-white mathematics of risk” (RR p.200).

Someone who learned to cross worlds fast. At the Project on Emerging Nanotechnologies, “I had to be an expert in everything, and I had to be able to build bridges fast”. He is “very un-disciplinary in what I do and how I do it”; being “stove-piped” is “putting the blinkers on”, and “being stuck in your discipline is actually an impediment” (TT p.2).

4. The moves#

4.1 Curiosity first#

Almost none of his pieces opens with a thesis. They start from something that caught him: the 2012 Indian blackout traced minute by minute (2015-01-30 responsible-development-of-new-technologies…), a student’s remark (2023-09-20 what-do-college-students-think-about-chatgpt), a Panama hat cracked in the Arizona sun (2026-02-08 beeswax-hallucinations-and-ai-inventions). His word for the pull is “intrigued”; surprise counts as data; small cases work as probes (“Despite the somewhat trivial example of using beeswax to repair a straw hat, it was clear that my epistemic vigilance has been well and truly circumvented”, 2026-02-08). He lets the question change as he goes: “This, I must confess, is not the question I started out with as I began working on this article” (2026-03-29 can-ai-create-an-undergraduate-degree-plan). And he turns curiosity on people: what the inventor in The Man in the White Suit lacks is “social curiosity”, the curiosity “to ask people what they think, and what they want” (FFTF p.222).

His curiosity is not naive. He traces the lure of permissionless innovation to it and confesses his PhD all-nighter: “Looking back, it’s shocking how quickly I sloughed off any sense of responsibility to get the data I needed” (FFTF p.161). Of a “maximally curious” AI: “I’m not sure there is a strong causal link between curiosity and benevolence” (2023-07-19 elon-musk-maximally-curious-agi).

4.2 Questioning the frame: reframing and first principles#

His most recognisable move is to take the term everyone uses unthinkingly, ask what it assumes and hides, and replace it with something closer to what matters. - “Risk aversion.” “I’m not sure I buy the idea of ‘risk aversion.’” It “can deflect attention away from what underlies many risk decisions: the things that people find too important to risk losing” (RR p.193). - Material definitions. A size cut-off is “a number of convenience, not of science”; “nature doesn’t care what we call a material, it just cares about how it behaves” (2022-02-10 are-we-asking-the-right-standards-questions…). - “Rogue” AI and extinction. “I must confess that here I get hung up by what is meant by ‘rogue’” (2023-05-25 leading-ai-expert-says-we-should). The extinction framing is “both too narrow and absolute a framing, and too human-centric” (2023-05-31 existential-risks-of-ai). - Optimist or pessimist? “It’s a bit like asking if I’m an oxygen pessimist or optimist” (2024-03-31 we-have-a-technology-problem-and). - Safety. From absolute to acceptable, on physical grounds: “zero risk — the corollary of absolute safety, is only possible in the absence of change” (2024-06-20 ilya-sutskevers-safe-superintelligence-rethink). - The “harness”. “metaphors are never completely neutral”; the danger is “treating the new as if it’s something old” (2026-02-22 what-we-miss-when-we-talk-about-ai-harnesses).

He reframes by going back to first principles: “it’s worth going back to first principles and what we mean by risk in the first place”, then five components beginning “no cause, no risk” (2023-11-26 everything-youve-heard-about-ai-risk-is-wrong); “Why are we considering developing standards related to advanced materials in the first place?” (2022-02-10); why learning matters at all (2025-03-30 reimagining-education-in-an-age-of-ai); what universities trade in: “We are part of a scarcity economy” (2026-09-24). And he flips questions: from what orphan risks are to “how they are made” (2026-07-16 orphan-risks-frontier-ai-maynard); from threats to the institution to threats to society, “you can flip this around in an interesting way” (2026-09-24); from AI’s accuracy to what it does for learning, which works “because of its limitations in some cases, rather than despite them” (2023-08-14 chatgpt-stimulates-creativity-critical-thinking).

His test of a frame is what it opens: “a framing of AI risks and benefits that opens up new possibilities rather than closing down conversations” (2023-05-31). Risk innovation itself began as such a reframe: “Imagine what might happen if we approach risk the way entrepreneurs approach innovation” (2016-01-11).

4.3 Structural analogy, with the break points marked#

He carries structure, not surface, between technologies, and marks where the carrying stops. - 2015. No single transmission line explains the Indian grid collapse; no single technology assessment prevents systemic failure. “The greatest challenge we face however, is in moving away from considering emerging technologies in isolation” (2015-01-30). - 2019, the clearest early worked example. He concedes, “At first blush, algorithms and hazardous chemicals have precious little in common”; states the limit, “Of course, an algorithm is not a chemical”; finds the analogy “intriguingly compelling” anyway; builds concepts by translation (is there an algorithmic equivalent of exposure? “I think there is”); tests them against real cars (“no clear algorithmic exposure route”), then adds “This in no way negates the research” (2019-03-05). - 2023–24. Nanotechnology’s governance debates as a mirror for AI (“Sound familiar?”, 2023-05-15 erik-schmidt-ai-regulation); Oppenheimer used mainly to show difference, AI’s risks being “hidden, dispersed, readily accessible” (2023-07-25 oppenheimer-and-ai); an 1889 dust counter revived because “seemingly novel challenges don’t always demand novel solutions” (2024-12-01 geoengineering-aerosol-monitoring-john-aitken, quoting his 2015 column).

The physicist polices the metaphors: “I am using this as a metaphor, no more” (2021-04-09); his “ideas mycelium” is “already pushing the metaphor farther than is probably wise, as any mycologist worth their salt will realize” (2023-08-21 the-messiness-of-the-provenance-of-ideas).

His use of analogy changes over time. In 2019 he translates with confidence. In 2024 he tries analogies out and demotes them: “I try and stay clear of analogies … But I’m going to go out on a limb this week … And then I’m going to explain why I don’t particularly like either of them”, keeping BlackBerry versus iPhone “not as a playbook … but as a mindset” (2024-05-05 blackberry-or-iphone-educational-ai). By 2026: “It’s rare that a new technology comes along which defies analogy with something we’re familiar with” (2026-01-22 think-you-know-ai-think-again). Even then he uses analogy as a probe and reads the break as the insight. Viruses are “both helpful and deeply unhelpful” as a comparison, because a virus “doesn’t instinctively know how to use every cognitive trick in the book to make us believe it’s alive” (2026-01-31 lost-in-the-moltbook-hall-of-mirrors). What carries over is how technologies meet society (“The specifics have changed enormously. The pattern hasn’t.”, 2026-04-12 self-account). What does not carry over is the category: “as soon as we start evaluating it within past frameworks, we make categorical errors” (2026-09-24).

4.4 Plausibility over imaginability#

The discipline that tightens the frame: “what is plausible, rather than simply imaginable, is vitally important” (FFTF p.171). He grounds it in chaos theory read for its limits: there are “boundaries to what might happen and what will not”, “highly relevant in separating out plausible futures from sheer fantasy”, and “points of stability”, with futures “that can be squandered if we don’t think ahead” (FFTF p.41).

He lets imagination run, then reins it in, and names this as his own technique: “a technique that I’ve used in the past to explore the plausible boundaries of emerging technologies, and one that works well for closing the shutters on hyperbolic speculation” (2024-11-17 navigating-the-ethical-dilemmas-of-brain-computer-interfaces). Examples: his 2014 3D-printed-brain thought experiment, labelled “a naive thought experiment” (2023-12-03 3d-artificial-brains-and-ai); road-death projections, “speculative as they are” (2016-04-01 will-driving-your-own-car-become…); a thermodynamic objection to superintelligence scenarios (2024-04-28 beyond-the-future-of-humanity-institute).

The test cuts both ways and can open as well as close. He keeps doubtful scenarios when they show a landscape’s shape: “Admittedly, some of these questions may lie beyond the realms of plausibility. But they do serve to highlight just how complex the ethical and governance landscape … is likely to become”. Of Neuralink’s slim chances, “Yet this is not the point here”: what matters is how reaching for such promises is “warping the pathway between where we are now, and the future” (2020-10-15 the-ethics-of-advanced-brain-machine-interfaces…). By 2026 this is a formula for deep uncertainty: “when the technology changes faster than we can generate data, you’ve got to have some degree of informed speculation, and some degree of imagination … don’t disallow speculation, but do it within a context of humility — knowing that it’s speculation, not reality; looking at possible futures rather than real futures; acknowledging that you need data to follow through; and bringing in different voices” (2026-09-24, n.4). In the same answer, AGI speculation is “blinkered and naive” and “nothing new under the sun” is “not evidence-based either”.

4.5 Building, experimenting and playing to think#

His strongest instinct is to find out by doing. - 2023. He runs ChatGPT through his annual review because “I felt obliged to see just how far this might be stretched” (2023-01-31 can-chatgpt-take-the-pain-out-of-annual-academic-reviews); runs escalating behaviour-prediction cases through GPT-4 (2023-05-22 can-large-language-models-be-used); reads over 2,000 conversations from his own course (2023-08-14). - 2024. He builds a Pippard’s ladder from Lego because “I was interested in whether it could be extended to thinking about different ways of approaching an uncertain future”. The four-quadrant framework came from “Experimenting with the ladder while thinking through the concept”; it was not designed first and illustrated after (2024-08-18). - 2025. Building a timeline of moral panics forced him to decide “what actually constitutes a moral techno-panic” (2025-06-01 vibe-coding-moral-panic). He reshaped a colleague’s diagram of AI discovery step by step: “What, though, if the spikes don’t all point in he same direction? … What if it’s spiky fractals all the way down?” (2025-07-27 spiky-surfaces-and-jagged-edges-moving). - 2026. He tests the AI-invented beeswax repair on the hat (2026-02-08); builds a 169-film corpus to test the dystopia trope (2026-05-15 ai-movies-may-be-less-dystopian-than-we-think); reports that his thinking about play was “influenced by actually playing the game — which is something I wasn’t expecting” (2026-08-02 what-we-can-learn-with-ai-by-not-trying-to-learn).

Four features recur. He is the instrument: in January 2023 it “intrigues me and slightly worries me that I’m sitting here already thinking of ChatGPT as a colleague and a collaborator” (2023-01-31); in 2024, playing a vulnerable user to a bot he built, “I was surprised at just how quickly it began to draw me in” (2024-10-27 personal-ai-chatbots-and-stochastic-agency). He calibrates: he tests AI “in an area where I would have a clear sense of where it was successful, and where it wasn’t” (2026-07-04 just-how-good-is-anthropics-fable-as-a-research-assistant), and deliberately stays a novice user to see what ordinary users can do (2026-03-29). He publishes the apparatus: prompts “typos and all” (2023-11-21 ai-and-risk-innovation), samples “admittedly with an ‘n’ of one” (2026-07-19 publish-or-perish-ai-vs-human-vs-human), failures included. He treats play as serious: “nothing is ever ‘just a game’” (FFTF p.221), and insight comes out of it: “I’m being a little playful. But in addressing these three points, intriguing possibilities do begin to emerge” (2025-07-27).

His play has a stated boundary. “Context is everything here”: experimenting “in a low-risk linear system where it’s relatively easy to turn the clock back” is one thing, but “I’d put breaking people, governance, society, and the planet, in this category!” (2025-03-02 the-lure-of-permissionless-innovation, n.2). This reversibility test squares a life of hands-on experiment with his critique of “move fast and break things”.

4.6 Serendipity, designed in#

He names serendipity early and arranges for it. In 2018: “I’ve been surprised and delighted at how these reflections have taken unexpected and serendipitous turns” (FFTF p.18). In 2015 his first worked example of risk innovation was “a book of seventeen haiku”, “an unusual result from an academic meeting”, set at one end of a spectrum whose other end is high-throughput toxicology (nnano.2015.196 p.731). Ideas serve as scaffolds: a colleague’s suggestion to write about the future as an object seemed “a little crazy”, then “formed a scaffold for what emerged rather than being central to the book” (2020-10-22 what-if-the-future-was-an-object).

He designs for it. Future Rising was “intentionally meant to stimulate the serendipitous emergence of new perspectives” (2021-04-09). His live conversations carry “absolutely no guarantee as to where we’ll end up going” (2023-09-18 will-ai-transform-how-we-learn). He paired two guests who had never met: “I intentionally set things up this way” (2024-03-15 liz-lerman-and-jonathon-keats-on). His Substack got a button so readers could “be randomly intrigued and delighted” (2025-04-20 surprised-by-serendipity). It is conditions, not luck: of his pizza seminar, “This, of course, wasn’t completely serendipitous”, because “I am very intentional in how I set up and guide the discussion each week” (2024-04-07 multigenerational-learning-tech-future). Unlikely sources are a requirement: creativity “means being willing to be influenced or inspired by unlikely sources. It’s one of the reasons I used Terry Pratchett” (2026-09-24, n.8). He defends serendipity as a public value too, asking whether we fund enough “exploratory and serendipitous science” (2024-10-08 ai-captures-this-years-nobel-prize).

4.7 Stories and films as thinking tools#

His use of film rests on an argument about risk. “Perhaps not surprisingly, risk is at the core of all the movies here. Each of these films has a risk-based narrative tension that keeps its audience hooked” (FFTF p.23). Stories are built from what characters stand to lose, “threats to dignity, belonging, identity, belief, even what it means to be human”, so they surface risks that hazard thinking misses: “I’m a sucker for using the imagination in science fiction movies to stimulate new ways of thinking about risk” (FFTF pp.23–24). For AI: “it’s plausible that some of these risks may blindside us, in part because we’re not thinking creatively enough about how an AI might threaten what’s important to us” (FFTF p.174). Creativity is a risk skill.

Films are a starting point, not a forecast: “a jumping-off point” (FFTF p.16), useful “precisely because they are not tethered to scientific accuracy” when “seasoned with feet-on-the-ground thinking” (FFTF p.288). Even a bad one works: “it’s the very absurdity of the movie that makes it useful” (2018-11-15 even-bad-sci-fi-movies…). Later he writes fiction as method, because of “affordances in fiction that allow complex ideas to be explored with a nuance and sophistication that all too easily elude more literal pieces” (2025-11-23 letters-from-the-department-of-intellectual-craft-prelude), and uses a fictional persona as an ethical experimental subject (2025-10-05 when-chatgpt-turns-snitch). Stories persuade where sermons fail: “Preach to someone about the future, and most people will shut down … But tell them a story that resonates with them, and you open their mind” (2024-01-21 how-can-stories-unlock-pathways-to). He also watches how stories mislead: technologists caught in a “Sci Fi feedback loop” (2024-09-18 neuralink-blindsight-brain-computer-interface), and the dystopia trope tested with data rather than repeated (2026-05-15).

4.8 Holding tensions, and navigating#

He holds opposites together: an obligation to innovate with its responsibilities (FFTF p.288); “Don’t Panic” with “we shouldn’t be complacent—far from it” (FFTF p.289); a “yes and” answer rather than either/or (FFTF p.269); “it would be foolish to discount emerging capabilities” (2026-06-12 a-quick-update-on-using-claude-fable-5) beside “one of the scariest things I’ve ever seen” (2026-09-24). These are not hedges. They are the ground to be crossed, and crossing it is what “navigating” means. Its intellectual basis is chaos with bounds: no control, but limits, stable points, and futures that can be lost (FFTF p.41).

So the method’s output is a map of pathways, not a ruling. Threats are to be understood “so that you can navigate around it … avoid it or flip it, and so get to the good” (2026-09-24, lecture body). “What the Planner does not do is provide answers to problems” (2023-11-21). To a research AI: “No recommendations at this point – remember the humility bit … But I do think you should be able to explore possible next steps and possible consequences of following certain pathways” (2025-02-04 openai-deep-research-ai-scholarship, his prompt). His models are disposable: “merely a thought experiment designed to stimulate new thinking. It may be so deeply flawed that it should be resigned to the trash can of bad ideas” (2024-08-18).

4.9 Humility as a working discipline#

• Against false precision: the “number of convenience” (2022-02-10); in his fiction, an academic outraged that his office shrank by exactly “eighty-nine square feet and twenty-three square inches”. Responsible science needs “the humility to recognize your limitations, and the willingness to listen to and work with others” (FFTF p.39).
• Uncertainty that grows with expertise: “Here, I freely admit that I may be wrong” (FFTF p.170). After thirty years in risk, “the less certain I am that we even know how to formulate the problems we face around AI”, and at the end of the same post, “Maybe I was a little hubristic with my statement” (2023-11-26).
• Revision in public: “I thought I was pretty savvy when it comes to navigating AI hallucinations. I was wrong” (2026-02-08, subtitle); “It’s a toy. It’ll never catch on.” “I was wrong.” (2026-09-24); “Re-reading this, I’m not so sure” (2026-04-26 why-im-falling-out-of-love-with-claude, n.3).
• Ways to be proved wrong: results that “would count against the incentive-driven account argued here” (2026-07-16); a pre-registered play experiment, because “where’s the fun — or the accountability — in that?” (2026-08-23 pre-registered-play-open-april-25).
• Declaring where he stands: “you should know a little about where I’m coming from, just so you can calibrate” (2026-09-24).

Humility never becomes paralysis. He acts on plausibility, and his 2026 formula ends with “bringing in different voices”, which makes humility collective.

4.10 Transdisciplinary by construction#

His self-portrait is Douglas Adams: “the skill with which Adams creatively melded together odds and ends of ideas from very different places to create new ones” (FFTF p.289). He pairs toxicology with machine learning, and cooperation biology with AI alignment (“one of those serendipitous moments that academics like me live for”, 2023-12-20 ai-superalignment-and-cooperation-science). What redeems brand-nano, for all his cynicism, is that it “broke down the barriers between previously stove-piped disciplines” (2018-02-21 the-bs-and-the-science-of-nanotechnology). He builds this into institutions: an NSF submission lists “the arts and humanities as potential modulators of advanced technology transitions” (2023-09-25 building-a-better-futures-tough); a colleague’s model of discovery becomes “n-dimensional space, where each dimension represents different scholarly and intellectual traditions” (2025-07-27). He is not against disciplines: the ideal keeps “much of the deep knowledge associated with traditional disciplines” while “transcending these disciplines” (2021-04-09).

5. How the moves fit together#

Together the moves form a cycle he runs again and again:

something catches him → he questions the frame → he loosens it (play, story, analogy, juxtaposition) → he tightens it (physics, evidence, first principles) → he builds or tests → he publishes, provisionally → he revises in public.

Underneath is the question of what matters to people; the goal is a way through. Three examples, seven years apart: - 2019-03-05, algorithms and chemicals. A headline catches him; he concedes the analogy looks odd, states its limit, translates exposure and dose–response into new concepts, tests them against how cars actually sense, holds the tension between zero and acceptable risk, and concludes a young field should import hard-won rigour “lest we end up building our understanding of algorithmic risks on an evidentiary stack of cards”. - 2024-08-18, the Lego ladder. A remembered demonstration; the question whether the metaphor can do more; building and playing; a framework emerging with mindset as an axis; “what if” turns (“But what if, instead of avoiding tipping points, we embraced them?”) that move him past his own earlier reading; precaution placed fairly as one stance among four; the whole offered as disposable. - 2026-02-08, beeswax. A trivial problem (“At this point, any sensible person would have spoken with a hat specialist … But of course I thought I’d go one better and ask Claude for advice”); being fooled “at the very moment I was writing about the risks of being suckered by Claude!”; the experiment run anyway; curiosity turning to the upside (“Had Claude inadvertently invented a new way to treat cracks in Panama hats?”); honesty about the result; the episode becoming evidence for his own epistemic-vigilance thesis.

The loosening and tightening are done by the same person within days. On 2021-03-28 he reasoned that released particles “up to around 5–10 µm in diameter could potentially present a health hazard”; twelve days later he lectured on odd-number universes and “metaphorical quantum tunneling” (2021-04-09). In March 2019 the dose–response case for algorithms (2019-03-05) appeared the day before the book chapter arguing that films help because they are untethered from accuracy (2019-03-06 navigating-the-risks-and-benefits-of-new-technologies). Each mode keeps the other honest.

The 2021 lecture explains why this combination suits technologies that break old categories. Conventional thinking is a “bounded infinity”: endless options inside a frame that excludes the ones we need, like a universe of odd numbers that can never reach an even one. Escaping it takes something like “metaphorical quantum tunneling”, and his chosen mechanism is “the juxtaposition of seemingly unrelated ideas” (2021-04-09). In 2025 the image returns for AI itself, as a possible “barrier-thinner” that lets us “tunnel” between separate fields of knowledge (2025-07-27). Play gets you out of the frame; physics and evidence make sure you land somewhere real.

6. How the method developed#

• 2014–17: evidence voice, creativity built in. Hazard versus exposure, hype deflated, and risk innovation founded on the claim that “this lack of creativity and flexibility in how potential risks are understood and addressed only increases the chances of things going wrong” (2016-01-11). Short stories used as risk theory (RR).
• 2018: the method written down. FFTF: films as lenses, plausibility, chaos with bounds, critical thinking plus creativity, “Don’t Panic”.
• 2019–22: teaching and serendipity. The film course, Future Rising as a “quantum tunneling enabler”, a switch from ethics commentary to mapping value (“Rather than writing another commentary on the ethical challenges …”, 2019-11-01 how-to-build-a-better-brain-machine-interface…), first playful AI experiments (2022-09-16; 2022-12-08).
• 2023: thinking in public. Rough posts because speed matters (“a little rough, but given the speed …”, 2023-04-04 what-are-the-alternatives-to-calling); self-experiment as early warning; reframes against pause, extinction and “rogue”.
• 2024: building and designed serendipity. Lego, headsets, conversation formats, playgrounds versus playpens; play named as the root of his method.
• 2025: instruments and fiction. Tools built to think with, models reshaped, fiction taken up, humility written into instructions to machines.
• 2026: self-implication and play defended. Analogy admitted to be failing, himself fooled on the page, joy defended (“I find ‘joy’ a deeply under-appreciated metric of intellectual and academic achievement!”, 2026-09-20, n.4), a formula for informed speculation.

The shifts: plausibility moves from mostly deflating fears to preparing for low-probability tails; analogy from confident translation to a probe of difference; play from practised to named; serendipity from accident to designed condition; and his position on AI from observer to participant-instrument. The register darkens; the method holds.

7. Nuances and tensions#

• Play and rigour. Enthusiasm sometimes outruns evidence (the dazzled early-2025 experiments; education claims resting on a course he designed). He usually flags it: “I may still be suffering from an overdose of ChatGPT and will come to my senses in a day or two. But I don’t think so” (2023-08-14).
• Curiosity as value and risk. He champions it and knows it drives permissionless innovation, including in himself (FFTF p.161). The reversibility test separates the two, but it is a judgement of context, not a rule.
• An analogy-maker facing a technology that “defies analogy”. His strongest tool strains, and he admits it: “even I am struggling to grapple with how to even describe what we are seeing” (2026-01-31).
• Shifting plausibility thresholds. The 2018 Occam’s-razor ranking of speculative futures and the later readiness to plan “on the off chance” rest on different thresholds. The 2026 formula is the nearest reconciliation.
• Himself as instrument. Designer, subject and judge are often one person. He names this and applies his epistemic-vigilance thesis to himself (“how do I know I’m not an unwitting victim here?”, 2026-01-17 i-cracked-and-wrote-an-academic-paper), but his safeguards are mainly transparency and readers’ scrutiny.
• The same levers for good and ill. He values stories, relationships and play because they get past defences, and fears AI because it does the same, possibly “beginning to train us to think like them” (2026-07-19). He has not yet offered a criterion for when such persuasion respects agency.
• Mental models or tools? Some concepts are also packaged as tools (the Planner, orphan-risk registers). His consistent line is mindset first, tools after.
• Taste and cost. He accepts that play-based methods “can be a matter of taste” (2026-09-20, n.1) and cost him standing: colleagues called his film book “professionally embarrassing” (2023-10-08 a-guide-to-responsible-innovation); the game “won’t do much for my academic standing” (2026-07-10). He persists, which shows how central the method is.

8. Connections to the other facets#

• Risk as a way of thinking. The method is why his risk concepts work as mental models. Threat to value, the risk landscape and orphan risks are reframes: they came from questioning “risk aversion” and “safe enough”, and from carrying innovation’s logic across to risk. Orphan risks are what imagination is needed to see (“not thinking creatively enough”, FFTF p.174). Navigation follows from chaos with bounds.
• What matters to him. The method is built to bring into view what people value, and it respects them: public resistance is read as protection of what matters, a reframe of “irrational” fear. Curiosity, joy and wonder are both his method and among the values he protects.
• Scholarship and public writing. Thinking in public is part of the method: writing is “integral to how I explore, test, and share new ideas” (2026-05-17 the-nonsense-i-write). Experiments go out with their apparatus; revision happens in view.
• His role as public scholar. The role is to pass the method on. Future Rising was designed to “catalyze metaphorical quantum tunneling in its readers” (2021-04-09). He builds playgrounds, offers games, questions and thought experiments rather than answers, and argues that “the joy of discovery” and “the freedom to ask questions that nobody else is asking” are what universities bring to the AI transition (2026-09-24).

9. What is distinctive, and what it brings to the AI discussion#

1. Two competences, each checking the other. Within twelve days he reasons about respirable particle sizes and about odd-number universes and quantum tunnelling as metaphor, using the physicist’s precision to police the physicist’s metaphors. Few voices in AI debates hold quantitative risk science and a practised imaginative method together.
2. Imagination as a risk competence. From 2016 he argued that a lack of creativity makes failure more likely, and from 2018 that AI risks will blindside us if we do not think “creatively enough”. Most risk assessment and AI-safety work treats imagination as speculation to be kept out. He treats it, disciplined by plausibility, as the way to see risks no framework yet owns.
3. Method before verdict, so no camp. He offers ways of seeing rather than positions, which is why he neither signs pause and extinction statements nor dismisses them. His refusal to polarise follows from how he thinks, not only from temperament.
4. Attention to frames. He reads debates for the assumptions inside their words (“rogue”, “extinction”, “tool”, “harness”), because frames lock thinking in. In a largely frame-driven AI debate, this is an unusual discipline.
5. Long memory without false continuity. He takes lessons about process from nanotechnology, GMOs, Asilomar and 1889 dust counters while insisting AI’s category is new (“defies analogy”; language as “the medium of formation”, 2026-09-24). This avoids both the “nothing new under the sun” dismissal and the opposite error of treating history as irrelevant.
6. An epistemics for technology that outruns data. “Informed speculation … within a context of humility” is a workable path between waiting for evidence that comes too late and treating speculation as fact.
7. Self-experiment as early warning. Relational pull and attachment, manipulation, “stochastic agency” and the evaluation gap were found by using the systems himself, usually before they became mainstream concerns, and he reports being fooled.
8. Play and serendipity as epistemic necessities, designed in. Playgrounds, unlikely pairings, rabbit holes and a serendipity button are his ways out of the “bounded infinities” of conventional thinking. For a world diverging from the conventional, this is the most distinctive part of his method, and the part the earlier concept map missed.
9. Humility shown, not claimed: public correction, falsifiable tests, pre-registration, “just so you can calibrate”.

What the method does not give is an operational procedure, and he says so (“no easy guidelines or rules of thumb”, FFTF p.39). From his perspective that is the point: in a domain nobody yet understands, a procedure would be false precision. What he offers is a way of thinking that keeps possibilities open while staying tied to reality.

10. The most revealing sources#

1. FFTF (2018), pp.16–26, 39–41, 161–174, 282, 288–291. The method written down.
2. nnano.2015.196 (2015) and RR (2017). Risk innovation as a culture of creativity and serendipity; a common term questioned; stories as risk theory; the physicist defeated by his own numbers.
3. 2021-04-09 bounded-infinities-quantum-tunneling-and-the-future-of-education. His theory of why juxtaposition and serendipity are needed, and his purpose as catalyst.
4. 2024-03-17 undergraduate-playgrounds-not-playpens. Play named as the root of his method, with the method at work in the same post.
5. 2019-03-05 should-we-be-treating-algorithms…, 2024-08-18 four-ways-of-thinking… and 2025-07-27 spiky-surfaces… Analogy with limits; thinking by making; reshaping a model to change what can be thought.
6. 2023-11-26 everything-youve-heard-about-ai-risk-is-wrong and 2023-05-31 existential-risks-of-ai. First principles, reframing that opens, humility that grows with expertise.
7. 2026-02-08 beeswax-hallucinations-and-ai-inventions. The whole cycle in miniature, self-implication included.
8. 2026-09-24 being-an-academic-in-an-age-of-ai (lecture body AI-smoothed; footnotes his), with TT (2023) pp.2–3 and 2026-09-20 reasoning-llms-just-want-to-have-fun. Flipping the lens, “categorical errors”, informed speculation, unlikely sources, delight, being “un-disciplinary”, and play, creativity and serendipity as method.