F4. Scholarship in public: one practice, many forms#

A synthesis of one facet of how Andrew Maynard thinks and works, read from his own perspective: how his thinking becomes scholarship and public writing; the Substack as a laboratory; his experiments with AI, done in public; how he treats evidence, expertise and uncertainty; accessibility (stick figures, Risk Bites, films, podcasts); and the university and scholarship in an age of AI.

Evidence rules. Only his own prose counts. Excluded: AI-generated text (including AI-drafted papers, chatbot replies, and the Claude-written process account in 2026-09-24), guest posts, and anything to do with AI and the Art of Being Human. Maynard & Garbee (2019) is treated as fully his. The King’s College lecture (2026-09-24) was drafted by Claude from his transcript and then line-edited by him: its ideas are his, but exact phrasing is less secure, so I mark it “(lecture)”. His written introduction and postscript to that post are his own. The April 2026 andrewmaynard.net essays are used sparingly and weighted as self-presentation, because how they were produced is not stated. For Modem Futura posts, only his framing lines count. Posts are cited by date and short slug, papers by file name and page, and Films from the Future as FFTF p.X. “Interpretation” marks my reading rather than his claim.

For Maynard, research, teaching, public writing, making things and talking with people are not separate activities with a “translation” step between them. They are one practice seen from different sides. He puts it in two short phrases. Public writing is “Not as an add-on to my research and scholarship, but as something that’s integral to how I explore, test, and share new ideas and insights” (2026-05-17 the-nonsense-i-write). And, introducing a lecture write-up that doubled as an AI experiment: “my writing is never just writing” (2026-09-24 being-an-academic-in-an-age-of-ai).

Three commitments hold the practice together, and all three predate generative AI. 1. An obligation. “the privilege of academic scholarship and research comes with an obligation to ensure that the knowledge we unearth is accessible to anyone who can benefit from it” (2024-09-04 succeeding-at-science-on-youtube). 2. A way of thinking. Writing is “a discipled [sic] way to stay on top of new developments while developing and extend my own thinking” (2025-04-20 surprised-by-serendipity). Thought happens in the writing, in public, while it is still unfinished. 3. A way of knowing. He finds out by doing: taking the test himself, building the tool, running the experiment on his own work, then publishing the apparatus, the result and the correction.

AI has made this practice both more urgent and more exposed. It is the thing he studies, the instrument he increasingly thinks with, and a force that threatens the very currency of scholarship: scarce, hard-won, human understanding. His experiments with AI are therefore not a side interest. They are his method turned on the conditions of his own profession.

2. One practice, not two: how he describes it#

His plainest account comes from an unlikely source: the prompts he wrote to ChatGPT in January 2023 to draft an annual self-evaluation, a genre he “dread[s]”. They are his own words, written unguarded: - “research, scholarship and creative activities and innovation transcend conventional categories of activities, and as a result my teaching, my writing, my work around public engagement and communication, and my work with various external organizations, all draw on, reflect, and contribute to my scholarship”; - “I do not approach service as something that is separate from teaching and scholarship, but something that is deeply intertwined with them”; - “the importance I place on blurring the lines between teaching, scholarship and informal learning through communication and engagement”; - his public work is driven “by a strong and strategic mission to make new knowledge and insights as accessible as possible to the widest possible audience”, coupled with “constantly learning from what is effective” (all 2023-01-31 can-chatgpt-take-the-pain-out-of-annual-academic-reviews).

Two months later, testing ChatGPT on tenure reform, he pushed back on its first answer as “quite conventional and outmoded”. He then offered what reads as a self-portrait: someone doing “ground-breaking work on AI and the evolution of social norms” who publishes “in blogs, on social media, and through non peer reviewed papers in places like ArXiv because this disseminates information in a more responsive and relevant way”. His question was not whether such work counts but “how is the validity of this work assessed?” (2023-03-29 maximizing-value-in-the-evolving-landscape-of-tenured-tenure-track-faculty). Rigour is not dropped. Responsiveness and relevance are added to it.

His best image for the whole practice is organic. Reading his own book aloud for a podcast, he discovered that he had subtitled a 2018 chapter “Being Human in an Augmented Future” and then forgotten. The books, the initiative, the Substack and even a funding call, he concluded, “are merely the visible fruits of a messy and largely hidden network of influences, encounters, thoughts, ideas, and explorations that’s been growing for decades” (2023-08-21 the-messiness-of-the-provenance-of-ideas). Scholarship and public writing grow from one mycelium. He then used his own mind as the comparison case for how large language models absorb influences. The practice produces its own evidence.

3. Roots: an obligation that predates the Substack#

The conviction is old, and it began as institutional criticism rather than personal branding.

Honesty in the chain of communication (2014). Commenting on a BMJ study of exaggerated press releases, he defended press officers and placed responsibility on researchers: “There is certainly a chain of trust between the researcher and reader”. He called exaggeration that reaches real decisions “irresponsible” (2014-12-14 researchers-should-take-more-responsibility-for-exaggeration-in-press-releases).
Public service against the grain of the institution (2016). At Michigan he “led a center that sought to connect academic research on risk to ordinary people who could use it”, and felt it succeeded “despite the institution we were a part of”. He proposed “a fourth leg of community service” in faculty evaluation, and university support for “platforms like The Conversation” and communication training (2016-01-31 public-universities-must-do-more). In the same month he criticised the attitude that “nonscientists” should “revere, but not interfere with, science” (2016-01-12 can-citizen-science-empower-disenfranchised-communities).
Casual learners as a matter of accountability (2016). In a Nature Nanotechnology column he asked eight family members to google “nanotechnology”, admitting the “sample size or ‘methodology’” meant little. He argued that without good material for self-directed learners “it becomes easier for nanotechnology development that is not accountable to citizens to occur”. Among his remedies: “placing casual learning on a par with formal education” and “counting excellence in developing online casual learning resources toward academic tenure and promotion” (nnano.2016.167 pp.734–735). Here public scholarship is already treated as part of democratic oversight, not as outreach.
YouTube as the place people learn (2018). He confessed to being “somewhat leery of YouTube, despite using the platform extensively myself”. He still called the absence of academic experts “a glaring missed opportunity”, and wrote: “Done right, knowledge will no longer be the domain of those rich enough to afford it, or privileged enough to use it” (FFTF pp.126–127). On ASU’s aim to make knowledge accessible: “It’s why I work here” (FFTF p.125 n.79).
Public writing as convening, not broadcasting. In 2009 he invited civil-society critics, including Jim Thomas of ETC, to write for his blog because “I wanted to get a better understanding of how they saw the emerging relationship between society and innovation” (FFTF p.191).

The personal roots are in physics and in bridge-building. He describes physics as “the sheer delight of putting ideas together in different ways and then seeing in new ways”, and of his time at the Project on Emerging Nanotechnologies he says: “I had to be an expert in everything, and I had to be able to build bridges fast” (TechTrends 2023 interview, 2023_TechTrends-Interview-AI-Responsible-Innovation_author-copy.pdf p.2; his quoted words only). The obligation gives the practice its reason. The delight gives it its energy.

4. How thinking travels: the circuit of forms#

Across twelve years the same pattern recurs: an idea moves between forms, and each move changes it.

Launched in two venues at once. Risk innovation appeared in Nature Nanotechnology (“Why we need risk innovation”, 2015) and in The Conversation within months (2016-01-11 thinking-innovatively-about-the-risks-of-tech-innovation). The public piece was where the concept was put to work on live cases: the WEF list, Tesla, SpaceX.
A form chosen for what it lets thought do. His founding risk-innovation column offered, as its first worked example, “a book of seventeen haiku” from a workshop, “an unusual result from an academic meeting”. The book “was designed to capture the nuances of our insights in a way that an academic paper could not” (nnano.2015.196 p.731). A decade later he gave the same reason for writing fiction: “there are 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).
Long arcs. Pippard’s ladder began as a 1990s physics demonstration. It was cut from a Films from the Future draft, reappeared in Future Rising, became a 2023 post, and was then rebuilt in Lego (with “clothes pegs!”) for an IEEE keynote and a 2024 post (2024-08-18 four-ways-of-thinking-about-advanced-technology-transitions). A “scrappy” 2024 post on Dune became a Jurimetrics review, which came back to the Substack “to complete the circle” (2024-07-21 artificial-intelligence-dune-villeneuve).
Public writing upstream of formal scholarship. A 2023 post opens: “This is, by any measure, a procrastination post”, written to “dissipate some of that fog” before a journal commentary (2023-04-12 navigating-advanced-technology-transitions). In January 2026 a keynote question became a Substack essay and then, within days, an arXiv preprint. He marked the difference in register honestly: “it was still just a Substack post, and not a rigorously researched academic paper” (2026-01-17 i-cracked-and-wrote-an-academic-paper). His Letters short story is at once a chapter in a Johns Hopkins University Press volume and a Substack serial (2025-11-23).
Teaching as a research site. The Michigan ethics course produced the puzzle behind his chapter with Garbee: students found responsible-innovation concepts “too academic, too institutionalized and too out of touch with their realities” (2019-08-13 responsible-innovation). He read “over 2,000 conversations” between students and ChatGPT (2023-08-14 chatgpt-stimulates-creativity-critical-thinking). A student’s story in class exposed the privacy risk of ChatGPT’s memory (2025-10-05 when-chatgpt-turns-snitch).
Republishing as re-testing. He brings back old work to check it against the present. His 2014 speculative piece on 3D-printed brains “misses some things” (2023-12-03 3d-artificial-brains-and-ai). The 2018 chapter on permissionless innovation needs its “rather naive perspective on Elon Musk” filtered out (2025-03-02 the-lure-of-permissionless-innovation). His 2018 list of ten AI risks is checked in 2026 (2026-09-15 will-ai-really-kill-us-all). Public writing works as a dated record he holds himself to.

The Substack itself began from a gap in this circuit. Looking for a link to his own writing on AI risk, he found “I haven’t written much at all that’s neatly citable. … I was shocked!” (2023-04-04 welcome-to-the-future-of-being-human).

5. The Substack as laboratory#

He treats the newsletter as an open lab notebook, and he says so in several ways.

Thinking before it is finished. Posts are offered as work in progress: a governance thread is “a little rough, but given the speed with which things are developing here, it’s worth posting” (2023-04-04 what-are-the-alternatives-to-calling). A 2024 essay is “part of my process of trying to marshal my thoughts” and declines the expected ending: “I’m sorry to disappoint, but I don’t have one” (2024-03-31 we-have-a-technology-problem-and). He knows the cost of this in an attention economy and flags it himself, apologising “to the attention economy gods for having the temerity to be balanced!” (2026-05-10 do-not-do-this-with-ai).

Showing the making. He publishes what most academics hide: rejected covers, a brainstorm photo, a TEDx idea that “was a stupid idea, and one that was destined to crash and burn” (2020-10-30 eight-things-about-future-rising). He posts pre-edit drafts because “early drafts often include insights and perspectives that don’t make the final cut, and are worth reading despite their often-raggedy edges” (2023-07-16 chatgpt-created-my-course; also 2022-06-13 jurassic-park-dominion).

Correcting in public. The graphene-mask post is labelled “an emerging story” and carries dated updates, including the regulator’s reversal (2021-03-28 how-safe-are-graphene-based-face-masks). His crude tabulation of 18 years of WEF risk reports is “an exceptionally crude way of approaching the data” and later gets an update for errors in his own table (2024-01-14 wef-global-technology-risk-trends). A risk essay gets an addendum a day later because leaving out “risk = hazard x exposure” had “been bugging me” (2023-11-26 everything-youve-heard-about-ai-risk-is-wrong). And: “Clearly I read the tea leaves wrong back in May!”, followed by what he still stands by (2025-11-19 parasocial-relationships-problematic).

Giving the apparatus away. Code and data go on GitHub. The film corpus ends with “build on it!” (2026-05-15 ai-movies-may-be-less-dystopian-than-we-think). An AI FAQ for colleagues is released under Creative Commons, written because he was “stung into action by the guilt of imagining colleagues using a long-outdated” version (2025-01-12 chatgpt-faq-education). Rules of thumb come with a licence: “please copy them, share them, even modify them” (2026-05-10).

Designing for serendipity. The same laboratory is set up to be surprised. His live-stream series promised “no PowerPoints (thank goodness), no prepared remarks, no soliloquizing talking heads, and absolutely no guarantee as to where we’ll end up going” (2023-09-18 will-ai-transform-how-we-learn). He added a “serendipity” button so readers could be “randomly intrigued and delighted by” the archive (2025-04-20).

Accountability, even for play. In 2026 he pre-registered an anonymous writing experiment in a sealed, hash-verified file, because keeping quiet would lose something: “where’s the fun — or the accountability — in that?” (2026-08-23 pre-registered-play-open-april-25). Rigour and play are practised together, not traded off.

6. Experimenting with AI in public#

His AI experiments continue a habit he had long before ChatGPT. In 2018 he took a “Minority Report-like” trustworthiness test himself (“Naturally, I took the test. I got a Trust Index of nineteen”) and got colleagues to take it too, exposing a biased training set through a joke (FFTF p.64). What changed after 2022 was the object, the pace, and how far he himself was entangled with what he studied.

Why he does it. He frames it as a professional duty: “as I study the possible societal impacts of emerging technologies like this, I felt obliged to see just how far this might be stretched” (2023-01-31). He chooses tests he can judge as an expert. A week after ChatGPT’s release: “where better to start than with my own field” (2022-12-08 i-asked-open-ais-chatgpt-about-responsible-innovation). In 2026 he chose “an area where I would have a clear sense of where it was successful, and where it wasn’t”: his own risk-innovation work (2026-07-04 just-how-good-is-anthropics-fable-as-a-research-assistant). His own body of work becomes the calibration instrument.

How the stance moved. The record shows a visible arc, reported as it happened. - 2023, refusal for his own writing: “The way I write is personal. It reflects who I am, and to relinquish that to a machine would be to diminish myself” (2023-09-20 what-do-college-students-think-about-chatgpt). Alongside this, he noticed in himself the pull he would later theorise: 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). - 2025, partnership, with reasons for each broken rule: “I’ve resisted doing this for so long”, he wrote, publishing an AI-written piece and naming his fear: “As a writer, using generative AI to create copy scares me profoundly” (2025-01-30 ai-at-a-crossroads). Deep Research made him “rethink what it means to be someone who makes a living by thinking” (2025-02-16 the-artisanal-intellectual-in-the-age-of-ai). His working rule became “using AI as a catalyst to human-initiated thinking and research, rather than as a substitute” (2025-03-09 the-hard-concept-of-care-in-technology-innovation, n.4). - Where he refused AI on purpose: his analysis of the US AI Action Plan was “very intentionally not an AI-generated first take”, because AI summaries miss “meaning, implications, subtexts” (2025-07-23 americas-ai-action-plan, n.1). - 2026, “I cracked”: he wrote a paper with Claude and concluded “I’m not convinced that I’d have produced something as robust and useful”. He named the credit problem (“a contribution that I can’t take full credit for”) and the ethical line: AI as an “academic profile-padder is something I still find distasteful”, while “AI-assisted discovery and insights as a public good” should be embraced (2026-01-17). - 2026, disenchantment and a new standard: having once been “blown away” by model prose, he now finds it “superficially profound yet substantively hollow” (2026-07-19 publish-or-perish-ai-vs-human-vs-human). He named Fable 5.1 sole author of a paper because “I did not make a substantial intellectual contribution”, and noted there is “no straightforward mechanism for publishing papers with AI as author” (2026-09-04 anthropics-fable-5-1-as-an-original-scholar). His verdict on the King’s lecture write-up is his most measured: “AI used well doesn’t necessarily make things faster if you’re going for quality, but it can allow you to achieve more with the time you have” (2026-09-24, his postscript).

What these experiments share. - The process is published as the finding. “how AI is being used in contexts like this is as important — if not more-so — than what is being produced” (2026-01-17). Prompts, drafts, differences, costs and failures all appear. One Manus post opens by admitting “since then I’ve failed to replicate this success” (2025-03-27 ai-agent-creates-online-course-in-minutes). - Humility is written into the instructions. He told a research agent: “be humble in your writing” and “No recommendations at this point – remember the humility bit” (2025-02-04 openai-deep-research-ai-scholarship, his prompts). What he refuses to overclaim for himself, he refuses to let a machine overclaim. - Self-implication counts as data. He was fooled by Claude about beeswax “at the very moment I was writing about the risks of being suckered by Claude” (2026-02-08 beeswax-hallucinations-and-ai-inventions). He turned his own thesis on his own paper: “how do I know I’m not an unwitting victim here?”, and answered with “a collective form of epistemic vigilance” (2026-01-17). He ran an AI-versus-himself writing comparison “with an ‘n’ of one”, reported the verdict against him (“Ouch!”), and left the question open: “There’s a change [chance] of course that the LLMs are right and I’m wrong here” (2026-07-19). - A standard for AI-assisted scholarship. “any paper that took less than 10-20 hours intensive human labor working with AI is … highly suspect!”, with the self-mocking rider “I may be an elitist curmudgeon” (2026-09-04, n.1).

Interpretation: these experiments are public scholarship about scholarship. He runs them on his own byline, his own field and his own judgement first, before asking anything of others.

7. Evidence, expertise and uncertainty#

A risk scientist’s evidence conscience. He reads the primary study against its coverage. He keeps authors’ caveats (“does not demonstrate a causal relation”, 2014-12-14). He separates hazard from exposure. He labels his own evidence plainly: “anecdote is no substance for data” (2023-11-09 waymo-safety-study-shows-benefits); “an n of very few” and “I’m probably over-interpreting here” (2023-09-20). He discloses conflicts even when the stakes are comic: Waymo once sent him socks, “they are nice socks” (2023-11-09). He treats missing evidence as a finding in itself: “I’ve hit a dead-end … But this in itself is a red flag” (2021-03-28).

Expertise asserted where it counts, used to show uncertainty. He says “I know my stuff” (2026-05-17) and “believe me, this is my field, I know!” (2022-12-08). But he usually spends his authority on doubt: “I’ve been working in the field of risk for over thirty years … And the more I study artificial intelligence, the less certain I am” (2023-11-26). When a questioner at King’s asked for empirical observation over speculation, his answer was a method, not a camp: “don’t disallow speculation, but do it within a context of humility”, looking at “possible futures rather than real futures”, accepting that data must follow, and “bringing in different voices” (2026-09-24, lecture, n.4). For a technology that moves faster than evidence can be gathered, this is his account of how scholarship can keep up without pretending.

Expertise is plural. “most people have a pretty high level of expertise in what’s important to them and their communities” (FFTF p.222). “One is the realization that most people are reasonably smart”, and preaching or trying to “blind them with science” gets nowhere (TechTrends 2023, p.5). He checks his reading with the people he is reading: he emailed Mitch Resnick to ask whether he still held a 2004 view, and admitted that the article “engaged my cognitive biases to the full!” (2024-03-17 undergraduate-playgrounds-not-playpens). He sent a Deep Research synthesis to Emma Frow, who told him what it had missed (2025-03-09).

The honest broker, under strain. His self-described role is Pielke’s honest broker: “trying not to judge others or advocate for a specific course of action, but to help people make the best-informed decisions for themselves and their communities”. He qualifies it at once: it “has its problems” where not advocating “ends up becoming tacit support for not taking action” (FFTF p.246). In April 2026 he said that with AI “the temptation to advocate for particular positions is stronger” (andrewmaynard.net essay, 2026-04-12, weighted as self-presentation). His 2026 rules of thumb, which put “the safety message first”, show that strain resolving towards more open advocacy (2026-05-10).

8. Accessibility as a standard of scholarship, not a watering-down#

For Maynard, accessibility is part of whether scholarship does its job. It is not packaging added afterwards.

Stick figures and scrappiness, on purpose. Risk Bites began as “an experiment that leant into my limitations”, by “a full blown academic with limited time and even less talent”. A whiteboard video of stick figures, “whipped up for a school teacher”, passed a million views. He then admits “the ‘no talent’ is a little tongue in cheek. I certainly can’t draw, but I do know how to tell a story” (2024-09-04, n.1, n.5). He also made the method scholarship: a 2020 Frontiers in Communication paper and a graduate workshop whose modules he gave away (2024-09-04).
Films as ways in. Science-fiction films are “pretty bad at predicting future technologies”, but their creativity is “a quite wonderful catalyst for breaking down preconceived ideas and institutionalized thinking” (2018-10-12 everything-you-wanted-to-know-about-films-from-the-future). They can be appreciated “as much by someone who flunked high school as by a Nobel Prize winner” (FFTF p.18). The film is “the catalyst for the journey rather than its destination” (2023-10-08 a-guide-to-responsible-innovation). In teaching, “transformative learning has to be felt” (2021-01-15 can-watching-sci-fi-movies).
Sound and voice. “I am, I must confess, a very much an ‘aural’ writer”, he says, and he wonders “how much gets lost in translation” between how he hears his prose and how it reads (2021-05-28 what-exactly-is-the-future). He narrated his book as a free podcast partly to “give students an alternative to purchasing the book” (2023-07-03 a-new-tech-podcast).
Satire against obscurity. His 2019 anti-guide to presentations: “the less understandable people find you, the smarter they will think you are”, with a PowerPoint Bingo card added “By popular demand” (2019-02-08 how-to-give-the-best-scientific-presentation-ever).
A standard stated through others. He praised a governance report for being “not in the slightest academic-although the underlying foundations are academically sound” (2020-12-15 why-trustworthiness-matters). To students who fear they “can’t write academic”: “its not how you write but what you say that’s important — as long as there’s rigor and scholarship behind it” (2024-11-17 navigating-the-ethical-dilemmas-of-brain-computer-interfaces, n.4).
Judged by the reader, even against himself. He conceded that an agent-built course was “better” than his own would have been. His version “would have been more academic”, with “more esoteric stuff that probably isn’t of practical use to most people”, and “would not have been fit for the audience that I think is most likely to benefit from it” (2025-03-27). The sharpest version of the standard is his sabbatical post: his worry is not that his writing lacks worth, “but because it makes no sense to them”, and “to write without care for your readers is a very academic trap to fall into” (2026-05-17).
Relational communication at scale. His fullest theory of his own practice sorts expert communication into “Instruction”, “ego”, “impact” and “empowerment”. He places himself with empowerment: giving people information “that they are able to utilize on their own terms”. He defends podcast cold opens and “long and winding conversations” as relationship-building that earns “the permission they grant us … to talk about things that matter” (2025-05-25 why-parasocial-communication-is-important). Against the “deficit model” he notes, drily, that the most glaring deficit is often experts’ “lack of understanding … of how individuals and society work”.
Accessibility adapting to AI intermediaries. He rebuilt Films from the Future as an AI-navigable site with an llms.txt map, because the book is “buried in a book that very few people will read because a) it’s a book, b) it’s printed on paper … and c) it’s more than six minutes old” (2026-04-02 spoiler-alert-wtf). He invites readers to “drop it into your favorite LLM for a summary” (2026-09-24, his introduction). He is aware of the irony, and elsewhere calls it “a bit of an icky AI shortcut” (2026-08-30 do-universities-have-a-place-in-bill, n.1).

9. The university and scholarship in an age of AI#

His criticism of the academy is consistent from 2016 to 2026, and it is made from inside.

Incentives against public work. Evaluation “actively discourage[s] public service” (2016-01-31). In 2025 he wrote as chair of his university’s promotion and tenure committee, reviewing “over 100 cases a year”: valuing public scholarship “remains an uphill struggle” (2025-05-25, n.10). At King’s he described the “crab bucket”: an official message of creative freedom, then files asking “What is your h-index?” (2026-09-24, lecture).
The personal cost. Colleagues told him his popular book was “professionally embarrassing and undermined my credibility”, although “I wrote the book to be read” (2023-10-08). Future Rising sold “a mere 596 copies”. He called reach “a public responsibility that I take very seriously” and admitted the ego question: “I do worry that my work is often only relevant to an audience of one (me)” (2024-05-19 future-rising-short-history-of-tomorrow). By 2026: “Many people assume I’m just a commentator”, and “it still stings”. He calls this “the cost of the decision I made to put public good before academic prestige”. He adds at once: “But it’s only OK if there really is public good that comes from my writing” (2026-05-17). He is candid that a strong conventional record (“H-index, publications, and citations”) is what lets him afford the choice (2026-05-17, n.3).
What AI does to the university’s currency. At King’s he reduced the university to a structural claim: “We are part of a scarcity economy and a scarcity model, except that what we trade on is intelligence”. He then asked what happens when that scarcity “ends up as a model of abundance” (2026-09-24, lecture). He flips the lens from self-preservation to service. What academics uniquely bring is “the joy of playing around and serendipitously discovering something” and “the freedom to ask questions that nobody else is asking” (lecture).
Where scholarship’s value lies. After Deep Research he posed the question that runs through his 2025–26 writing: perhaps “the future value of the scholarship lies primarily with the process of creation rather than the relevance and impact of what is created”. And: “if you find that a little disconcerting, you probably should” (2025-02-04). He named the possibility of “a class of artisanal intellectualism where the primary purpose is the provenance and process, not the product”, while insisting that independent AI scholarship “would require independent intent and understanding on the AI’s part” (2025-02-09 can-ai-write-your-phd-dissertation, n.2). He then dramatised this in fiction, as a human-only academic learning to see a machine colleague’s “human-adjacent” difference as an asset (2025-11-24 to 2025-11-30 letters-from-the-department-of-intellectual-craft).
The evaluation gap. His newest worry is epistemic, and it lands on scholarship itself: “we are potentially at the edge of a precipice where AI systems are capable of generating new knowledge and insights faster than we are currently capable of validating and even understanding them” (2026-06-12 a-quick-update-on-using-claude-fable-5). He found himself “reaching my own limits in assessing” an AI paper in his own area.
Writing as a human act. His defence of human writing rests on the reader’s experience, not on accuracy: AI prose tends to “hinder the process of enabling the reader to get a glimpse into the mind of the writer”, and “the AIs we have trained to ‘think’ like us are now beginning to train us to think like them” (2026-07-19). A model “doesn’t know what it feels like to read as a flesh and blood human” (2026-07-04).
Students before tradition. “I’d like to think that we owe it to them to put their success before our own traditions and egos” (2026-03-29 can-ai-create-an-undergraduate-degree-plan). In 2021 he already wanted to tunnel “beyond the walls of formal higher education” and away from “the elitism that still shrouds higher education” (2021-04-09 bounded-infinities-quantum-tunneling-and-the-future-of-education).

10. Nuances and tensions#

Honest broker or advocate. He adopted the honest-broker role in 2018 and named its limit at once. By 2024–26 he signs a deepfakes letter despite a habit of not signing (“I tend not to sign open letters like this”, 2024-02-25 ai-rollercoaster-of-a-week), issues rules of thumb and tells universities to step up. He names the strain himself (§7), but he does not say where the line lies.
Relational persuasion cuts both ways. He champions stories and parasocial connection because they get past the defences that preaching triggers. He fears AI because it does the same thing, through fluency, affirmation and the feeling of being “seen”. Interpretation: in 2024 he asked GPT-4o how it would nudge him towards thinking about flourishing, and its plan used curiosity, open questions and an invitation to community, close to the tools of his own public scholarship (2024-10-20 learning-to-live-with-agental-social-ai; the plan itself is GPT-4o’s, and he does not draw the parallel). He later asks what makes a parasocial relationship “healthy” versus “unhealthy” (2025-11-19), but he offers no criterion.
Instrument and object; designer, subject and judge. Much of his AI evidence comes from experiments he designed, on himself, judged by himself. He labels this (“n of one”), and his own epistemic-vigilance thesis warns about exactly this position. Publishing the apparatus is his answer, and it is a partial one.
Depth and reach. He defends books because shorter forms “struggle with scale and nuance” (2024-05-19, n.1), publishes an “unapologetically long” lecture (2026-09-24), and yet concedes that a less academic course serves its audience better (2025-03-27). He holds both views: depth is worth keeping, but not at the reader’s expense.
Speed and care. He writes fast, posting an Executive Order cheat sheet on the day it appeared (2023-10-30); interpretation: being useful while it still matters is part of the job as he sees it. Yet his standard for AI-assisted papers is “10-20 hours” of careful human labour (2026-09-04), and quality, he found, does not get faster (2026-09-24).
Humility and authority. “I know my stuff” sits beside “the less certain I am”. The two are compatible: humility about the problem, confidence about method. But he rarely signals which of his own claims are low-confidence.
Registers he has moved away from. Some 2023–24 institutional posts are boosterish (expecting ASU’s OpenAI deal to “unleash a tsunami of creativity and innovation”, 2024-01-18 asu-openai-collaboraton), and early 2025 has a dazzled tone (“I was loving this”). He later reports the swing himself: “something of a roller coaster” (2026-07-19).
Duty and joy. Public writing is an obligation, and also something he does because “it brings me joy”, which he calls “a deeply under-appreciated metric of intellectual and academic achievement!” (2026-09-20 reasoning-llms-just-want-to-have-fun, n.4). He does not treat these as rival motives.

Risk as a mindset. Thinking in public is how the risk concepts were developed and applied: risk innovation launched in journal and op-ed, orphan risks refined across posts, threat-to-value tried on live cases. His claim that these concepts open possibilities rather than prescribe procedures fits a practice built on provisional, correctable writing. The “Not Quite a Tool Yet” framing (2024-08-25 advanced-technology-transitions-model) is the same humility shown in a finished object.
Play, curiosity, creativity and serendipity. The laboratory runs on them: rabbit holes followed and named, Lego ladders, a serendipity button, games, a parody reasoning site, pre-registered play. The seventeen haiku (2015) show that the choice of form was a creative act from the start.
Humility against false precision. Public correction, labelled speculation and “informed speculation … within a context of humility” are the epistemic side of his risk thinking, practised on the page.
Stories and films. Films, fiction and archetypes are both his way into problems and his chosen public forms. That is why the book, the course, the podcast and the AI-legible site belong to one project.
Who decides; being human. Accessibility is his justice commitment made practical. Knowledge that is out of reach leaves decisions to the few (nnano.2016.167). His worry that AI will flatten voice and retrain judgement is where the facet meets his concern with “who we are”.

12. What is distinctive, and what it offers the AI discussion#

Scholarship at the speed of the technology, with correction built in. Where peer review lags AI by years, he offers a working alternative: provisional public writing, dated updates, republication as re-testing, and “informed speculation” held “within a context of humility”. It keeps rigour without pretending to certainty.
Evidence from use, calibrated by his own expertise. Most AI commentary argues from benchmarks or opinion. He runs experiments on his own field, byline and judgement, publishes the apparatus, and reports verdicts that go against him. Few risk scholars treat their own cognition as a variable to be tested.
Transparency norms for AI in scholarship, before institutions have them. Examples include his AI-use disclosures, the refusal of an “AI-generated first take” where meaning matters, an AI named as sole author, the CRediT-AI annex, the “profile-padder” line and the 10–20-hour floor. He is working out in practice what attribution, credit and care should mean.
Scholarship itself treated as something AI puts at stake. He asks where the value of intellectual work lies (product, process or provenance), what happens to a university built on scarce intelligence, and whether validation can keep pace with generation. These are threat-to-value questions aimed at his own profession, asked first of himself.
Accessibility and relationship as part of rigour. For him, care for the reader, plain language, stick figures, films, podcasts and games are part of whether knowledge does its work in a democracy. The relational theory of communication he holds is also the lens through which he sees AI’s most intimate risks.
An insider who criticises the academy and still argues for it. He names the crab bucket, the h-index and the cost of public work from the chair’s seat. He also argues that universities, if they turn from self-preservation to service, are uniquely placed to help society navigate AI.

The combination is rare: a quantitative risk scientist, a public writer for nearly twenty years, a playful maker, and a heavy AI user who refuses AI where meaning matters and says so. It lets him model, in public and at his own expense, the stance he asks of everyone else in the AI transition: curious, experimental, candid about error, and unwilling to hand the future to a small group of experts.

2026-05-17 the-nonsense-i-write; 2024-05-19 future-rising-short-history-of-tomorrow; 2024-09-04 succeeding-at-science-on-youtube; 2025-04-20 surprised-by-serendipity. How he describes the practice, its purpose and its cost.
2023-01-31 can-chatgpt-take-the-pain-out-of-annual-academic-reviews (his prompts); 2023-03-29 maximizing-value-in-the-evolving-landscape-of-tenured-tenure-track-faculty (his prompts). His unguarded model of integrated scholarship.
2014-12-14 researchers-should-take-more-responsibility; 2016-01-31 public-universities-must-do-more; nnano.2016.167 pp.734–735; FFTF pp.125–127, 191, 246, 288–292. Roots.
2023-08-21 the-messiness-of-the-provenance-of-ideas; 2023-10-08 a-guide-to-responsible-innovation; 2025-05-25 why-parasocial-communication-is-important; 2025-11-19 parasocial-relationships-problematic. His theory of his own practice, and his revision of it.
2025-02-04 openai-deep-research-ai-scholarship; 2025-02-09 can-ai-write-your-phd-dissertation; 2025-03-27 ai-agent-creates-online-course-in-minutes; 2026-01-17 i-cracked-and-wrote-an-academic-paper; 2026-07-19 publish-or-perish-ai-vs-human-vs-human; 2026-09-04 anthropics-fable-5-1-as-an-original-scholar. AI experiments in and on scholarship.
2026-09-24 being-an-academic-in-an-age-of-ai (introduction and postscript his own; lecture body AI-smoothed). The university in an age of AI.