Late Lessons, Jensen Huang and AI
Markdown version

SUPP-self perspective notes: how Andrew Maynard describes his own thinking (self-accounts, 2011–2026)#

Scope. This batch covers the texts where Andrew explains himself: what he does, how he does it and why. It also covers the Nature and Nature Nanotechnology columns that show where his method came from, and the stated mission of the Future of Being Human initiative. I read everything below in full. I did not open files 01–04 or 06, working/huang, working/late-lessons, working/synthesis or working/maynard-lens.

What was read, and how much weight it carries.

Citation key. - FR = Future Rising (2020), page numbers from the printed page headers. - RR = Rethinking Risk, book page numbers. - TT = TechTrends 2023, PDF page. - Nature Nanotechnology columns are cited by file name and journal page, for example nnano.2015.196 p.731. - Web essays are cited by date and short file name. Posts are cited by date and slug.

1. How he thinks#

He starts small and concrete, then sees the pattern. Almost every self-account opens with a scene rather than a thesis: - a packet of carbon nanotubes being opened at NIOSH in 2004; - a CAT-scan consent form; - a Congressional hearing room; - poison ivy; - Michigan potholes; - a white square of fabric in the National Gallery; - an email from a stranger worried about a fire extinguisher; - his family googling “nanotechnology” for him.

From the scene he moves to the structure. The nanotube question of 2004 and the AI question of 2026 are “structurally identical”. Then comes the line he uses to connect thirty years: “The specifics have changed enormously. The pattern hasn’t.” (2026-04-12 What-Thirty-Years). The pattern is a gap between what a technology can do and a society’s capacity to handle it. He insists “the gap is never primarily technical” (same essay).

He thinks like a physicist, and the physics is about delight, not procedure. In his own words: - “physics is all about the sheer delight of putting ideas together in different ways and then seeing in new ways. I’ve never lost that delight” (TT p.2). - “Science is a love language between us and the universe” (TT p.2). He adds that the way science is now taught as “a process or a method” “destroys my soul”. - Future Rising builds a humanistic argument about responsibility to the future out of entropy, light, causality and complexity (FR pp.25–77). Physics gives him metaphors and a way of seeing, not only numbers.

He reframes by questioning the word everyone takes for granted. This may be the most recognisable move in these texts. He takes a term in wide use, asks what it assumes and what it hides, and replaces the category with something closer to what actually matters: - “I’m not sure I buy the idea of ‘risk aversion’” (RR p.193). Beneath it lies “the things that people find too important to risk losing”. - Don’t define nanomaterials: a size-based definition is “a ‘term of art’ rather than science” (Nature 475:31, 2011). - Is novelty overrated? Novelty “favours the interesting (and possibly the headline-grabbing) over the important” (nnano.2014.116 p.410). - “‘Nanotechnology’ is an unreliable indicator of risk” (nnano.2016.28 p.211). - The “harness” metaphor for AI. He says he is “pulling apart the language we use to talk about AI”, and asks for “some intentionality around the framing before it locks in” (2026-04-12 Honest-Non-Signals). - The Sorcerer’s Apprentice, “and not in the way it’s usually invoked” (2026-04-12 Three-S-Curves). - The “consciousness trap” (2026-04-12 Future-Were-Building).

He draws on unlikely sources on purpose. His examples include: - two short stories as a way into risk theory (RR); - Terry Pratchett’s crab bucket and Malcolm Bradbury’s campus novel in a lecture on universities and AI (2026-09-24); - the Vatican’s Antiqua et Nova read side by side with the International AI Safety Report (2026-04-12 Future-Were-Building); - Paul the octopus and a flat-earther’s rocket (FR), and a Flanders and Swann song (2026-09-24).

After the King’s College talk he said that creativity “means being willing to be influenced or inspired by unlikely sources”, which is “one of the reasons I used Terry Pratchett” (2026-09-24, fn 8; AI-edited transcript). The acknowledgement in Rethinking Risk describes the same method: “serendipitous insights that come from ‘yes and’ collaborations between creative writers and technical experts”, and thanks the authors for “helping me see the world I thought I knew through new eyes” (RR p.200).

He speculates, labels it as speculation, and anchors it to what is plausible. He will run a thought experiment and call it one: - “Of course, this is a naive thought experiment” (nnano.2014.294 p.956). - “this admittedly speculative assessment” (nnano.2015.286 p.1006). - “some of this could be completely wrong… These are explorations, not findings” (2026-04-12 Honest-Non-Signals).

The discipline comes from the risk scientist’s habit of benchmarking, which “helps tether speculative ideas to plausible realities” (nnano.2016.28 p.211). His clearest statement of the balance came in answer to a techno-optimist at King’s College: “don’t disallow speculation, but do it within a context of humility”. He glossed that as knowing it is speculation, “looking at possible futures rather than real futures”, accepting that data must follow, and “bringing in different voices” (2026-09-24, fn 4/5; AI-edited).

He tests things by doing them, and reports his own misjudgements. Examples: - He co-designed a prompt-engineering course with ChatGPT, expecting he “could sit back and do nothing. Boy, was I so wrong” (TT p.3). - He had ChatGPT Deep Research write a 400-page dissertation, and was careful to say “I’m not sharing that to be alarmist” (2026-04-12 What-Thirty-Years). - He built a Character.AI bot and then presented to it as a vulnerable user (2026-04-12 Honest-Non-Signals). - He rebuilt Films from the Future as an AI-navigable site. - He added a “serendipity button” to the Substack (2025-04-20). - He “pre-registered” an anonymous writing experiment in a sealed file, because keeping quiet would lose “the fun — or the accountability” (2026-08-23). - On early GPT APIs: “It’s a toy. It’ll never catch on.” Then: “I was wrong.” (2026-09-24).

He changes his mind in public and treats it as fuel. - “Five years ago, I was a proponent of a regulatory definition of engineered nanomaterials. I have changed my mind.” (Nature 475:31, 2011) - Of fumed silica: “here was evidence that cast doubt on what I thought I knew to be true” (nnano.2014.196 p.658). He weighed it and still judged the material acceptably safe. What changed was how certain he was, not his conclusion. - “When I discover I’m wrong… I find it amazing, and that becomes fuel to my creativity” (TT pp.2–3). - Exploring edutainment, he was “willing to have my mind changed”. He also admits a source “engaged my cognitive biases to the full” (2024-03-17). He is candid about confirmation bias even when being candid about openness.

He holds tensions rather than resolving them early. - Hubris both drives and blinds: “Our hubris pushes us forward”, but it can “blind us to the realities of what’s possible” (FR p.150). - On AI: “stuck between thinking this is one of the scariest things I’ve ever seen… and… that the potential is profound… I’m neither an AI optimist nor an AI pessimist. But I do have really, really bad days” (2026-09-24; AI-edited). - On fumed silica he sets out three kinds of irresponsibility: questioning its use on thin evidence, curbing research, and failing to explain the difference (nnano.2014.196 p.659).

Play and serendipity are the working method, not decoration. His clearest self-description is in a postscript: “So much of how I explore new ideas, put knowledge and understanding together in different ways, and revel in the serendipity of new discoveries, is grounded in play.” He traces this to undergraduate physics labs, where “the tactile problem solving and creative exploration… fired my imagination” (2024-03-17). The same post shows the method at work: - a “research ‘methodology’” of playing around on Google; - then emailing Mitch Resnick to ask whether he still held a 2004 view; - then arriving at the playpen-and-playground metaphor, which became a lasting frame.

Future Rising itself came from a friend’s “very off question: ‘why don’t you write about the future as an object?’”, and “within 24 hours I’d gone from skeptic to enthusiast”. The idea was a scaffold he later took down: the book “turned out not to be” about that (FR excerpts cover note, pp.1–3). Writing the book, he found “moments of serendipity” and was “surprised, delighted, and humbled” (FR p.215).

2. What matters to him#

3. Risk as a way of thinking#

The foundation is quantitative, and he builds on it rather than discarding it. He began as an aerosol physicist measuring nanoparticle exposures at the UK Health and Safety Executive and at NIOSH. His career includes work in the early 1990s collecting and imaging particles “just a few nanometres in diameter” (nnano.2015.120 p.482). - He respects what numbers do: “Probability as a numeric representation of risk is a powerful way of making trade-offs” (RR p.193). - Calculus is the “language of change” that helps “identify impending dangers, as change threatens to take away what we value”. It works “when tempered with humility and guided by our humanity” (FR pp.166–167). - Benchmarking, exposure metrics, trigger points and the NIOSH exposure limit all appear as live tools in his columns (nnano.2016.28, nnano.2016.99). - When he first defined risk innovation, one end of its range was purely technical: the EPA/NIH Tox21 high-throughput toxicology programme (nnano.2015.196 p.731). - He calls threat-to-value “an evolution of the old black-and-white mathematics of risk”, an evolution and not a replacement (RR p.200).

Why the numbers are not enough: the CAT-scan story. Faced with a one-in-a-million chance of serious harm from contrast dye, he writes: “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… but because that was what I was expected to do”. The value that decided it was not embarrassing himself (RR p.195). Two lessons follow from his own body: - numbers “can be deceptive”; - risk calculations depend on “who decides what’s important” (RR p.194).

Risk as a threat to value. “Risk starts with something that is worth protecting.” Worth includes health and environment, and also “security, friendships, social acceptance, and our sense of personal and cultural identity” (nnano.2016.28 p.211). Risk innovation “frames risk as a threat to existing or future ‘value’” (nnano.2015.196 p.731). He is clear that this changes what risk is for: “Risk in this instance is not a danger to be avoided, but an inevitability that reveals what the primary value is within a complex landscape” (RR p.197).

Landscape and navigation, not management. In his titles he navigates rather than manages: Navigating the risk landscape (2016), Navigating the fourth industrial revolution (2015). The picture is geographical: - “a much larger and murkier risk landscape” (nnano.2015.196 p.730); - “shifting hills and valleys” (nnano.2016.28 p.211); - guideposts to “plot a sensible course”; - learning to navigate “albeit slowly, and sometimes hesitantly” (nnano.2016.28 pp.211–212).

The landscape “cannot be navigated without a similarly complex and multidimensional understanding of risk” (nnano.2015.196 p.730). In the 2026 retrospective, risk innovation is “a reframing of risk from something to be minimized to something to be navigated creatively in pursuit of value” (2026-04-12 What-Thirty-Years).

Risk as opening possibilities. The value frame lets risk conversations “be elevated from simplistic ‘go/no-go’ options” to how gains and losses are balanced across everyone affected, and “this elevation in turn opens the door to creative and innovative approaches to protecting existing and future value” (RR pp.197–198). Other statements of the same idea: - Risk is “not only endemic, but integral to progress” (nnano.2015.196 p.731). - In 2026: “you can only begin to realize the benefits of a technology if you understand what can possibly go wrong, so that you can navigate around it… avoid it or flip it, and so get to the good” (2026-09-24; AI-edited). - The TechTrends version: “a delight and a real importance” in asking with stakeholders “what could possibly go wrong, and how we can get it right” (TT p.2).

Creativity, serendipity and play were there from the start of risk innovation. This is the strongest documentary support for how he now describes himself. The 2015 founding statement: - calls for “parallel innovation in how we conceptualize risk”; - gives “license to what might be described as risk entrepreneurship, where the ultimate measure of an idea’s worth is whether it has an impact, not whether it adheres to convention”; - asks for “a culture of experimentation — a culture grounded in transdisciplinarity, creativity and imagination; and epitomized by serendipity and a ‘fail fast fail forward’ mentality” (nnano.2015.196 p.731).

His first worked example is not a model. It is “a book of seventeen haiku” from a workshop with the V2_ Institute for the Unstable Media, “an unusual result from an academic meeting”, meant to open “risk navigation pathways… that would otherwise remain hidden” (p.731). He places art and Tox21 at the two ends of a single range.

Why conventional risk thinking must change. Well before AI he argued that established methods were “developed as a consequence of, the impacts of previous industrial revolutions”. Converging technologies need us “to be jolted out of our existing mental and procedural risk-ruts” (nnano.2015.286 p.1006). He applied the same critique to his own field: - The Royal Society/Royal Academy of Engineering 2004 report set risk research on a path that “has, over time, worn a rut”. “The speculation of possible risk has developed into an assumption of as-yet-to-be-discovered risk.” - He asked whether researchers had “created a new, metaphorical grey goo” (nnano.2014.43 p.160).

By 2026 the argument is sharper for AI: “as soon as we start evaluating it within past frameworks, we make categorical errors” (2026-09-24; AI-edited).

The things no framework owns. Orphan risks are threats to “dignity, belonging, identity, autonomy, democratic participation, what it means to be human” that “no existing institution owns” (2026-04-12 What-Nanotechnology). He also defends “mundane” risks against both hype and dismissal: “mundane risks are still risks” (nnano.2014.116 p.410).

Early warnings, trigger points and timing. This view is nuanced and is not an anti-precaution stance. - “We need to be quick to question, and slow to respond”. Yet we “also need the ability to respond proactively where early warnings of potential harm do begin to emerge — even before the science is mature” (nnano.2016.28 p.212). - He asks “how are appropriate trigger points for action defined?” and warns that “premature action could cause a lot more harm than good” (nnano.2014.196 p.659). - In 2011 he proposed adaptive, material-specific “trigger points” in place of a black-and-white definition. He cited Libby vermiculite as a harmful mineral that “slipped through the regulatory net” because of a definition (Nature 475:31). - The through-line is to keep “what if” research free, and to be clear about when evidence becomes actionable.

Humility against false precision. - “The more precise we try to be with our predictions of the future, the less likely they are to be accurate”. The danger is to “act as if the future is something we can fully control” (FR p.148). - The War on Cancer is his example of hubris. “Just because we are smart and can imagine a future… doesn’t mean that it’s automatically within our grasp” (FR p.150). - “There are even indications that the smarter we are, the better we are at justifying our beliefs in spite of evidence” (FR p.153). This anticipates his 2026 “intelligent user trap”. - For AI: “we don’t have theories of advanced technology transitions. We don’t have a body of knowledge that helps us understand how to actually navigate this successfully” (TT p.5). - “How you think about nanotechnology risk is probably incomplete”, so collaborators need “the humility to recognize and respect expertise outside of their domain” (nnano.2016.28 p.212).

How this fits his own September 2026 account. The self-texts strongly support his account, with three refinements. 1. The call to change mindset has two layers. The general layer, argued since 2014–15, is that emerging and converging technologies outrun risk frames built for earlier industrial revolutions. The AI layer, argued since 2022–24, is that conversational AI is the first technology to enter human self-constitution. It changes who we are, not just what we can do (2024-01-01; 2026-04-12 Future-Were-Building, Honest-Non-Signals). 2. He holds continuity and novelty together. He argues that frames must change. He also argues: - “seemingly novel challenges don’t always demand novel solutions” (nnano.2015.120 p.483); - novelty is an unreliable guide to risk (nnano.2014.116); - the pattern of the gap “hasn’t” changed (2026).

What carries over from history is the human and institutional pattern, the lessons about process, and humility. What does not carry over is the categorical frame: definitions, hazard lists, novelty as a proxy. 3. The concepts work as mental models, not procedures. Even his practical tools are offered to open thinking. The 2016 guideposts are “a starting point”; the 2023 Risk Innovation Planner “might have taken the edge off some of the messiness” (2026-04-12 What-Nanotechnology). The initiative’s value of “grounded exuberance”, pushing “at the boundaries of conventional thinking while remaining grounded in reality” (FoBH homepage), is the risk scientist’s balance put into the language of play.

4. Scholarship and public writing#

Writing is how he thinks. - He gives three aims for public writing. The second is “using writing as a discipled [sic] way to stay on top of new developments while developing and extend my own thinking” (2025-04-20). - On the podcast: “the conversations that Sean and I have are part of our own intellectual journeys… places where we get to think about and test new ideas… and to be delighted by the serendipity of unexpected insights”. These are journeys “our listeners are invited to join us on” rather than being kept in “the inaccessible corridors of academia” (2024-12-22). - He started the Substack after failing to find anything “neatly citable” of his own on AI risk. That was “the jolt I needed” to return to informal, responsive writing (2023-04-04).

One idea travels through many formats. In his own tracing: - a conference provocation in Berlin became an arXiv preprint (2026-04-12 Honest-Non-Signals); - Risk Bites videos became a Frontiers in Communication paper and a training programme (2026-04-12 Stick-Figures); - thought experiments stay as Substack essays that he holds “more tentatively” (2026-04-12 Honest-Non-Signals); - a 2018 Risk Bites list of ten AI risks was updated in 2023 and still held, because “the risks I focused on were human risks” (2026-04-12 Stick-Figures).

He describes each channel as reaching a different public: academics, Substack readers, casual YouTube learners, feed-driven readers (same essay).

He writes the same way in Nature as on the Substack. His Nature Nanotechnology columns use the first person. They confess, tell family stories and joke: - his eight “volunteers” were “immediate family members who kindly googled nanotechnology for me” (nnano.2016.167 p.734); - his son’s hackathon (nnano.2015.35 p.200); - the MWNT-7 nanotube, named after an employee’s “Lucky-7” shipping number, “if, ironically, not the luckiest” (nnano.2016.99 p.491); - a student asking “can we trust papers more than a few years old?” (nnano.2015.120 p.483).

The scholarship and the public voice are one voice.

Stories and art are serious tools. - Short stories “help unpack the concept of risk from the perspective of what’s important to whom” (RR p.199). - “Art, in all its forms, is oxygen to our creativity”; it works by “kicking us out of the rut of conventional thinking” (FR p.96). - Science fiction films “break down barriers between experts and non-experts, bypass ideological divisions, and give everyone permission to think about the future”. Framed as a policy paper, the same conversation “would clear the room”. Movies are “democratic entry points” (2026-04-12 Stick-Figures). - A caution he keeps alongside: “Make-believe treated as reality has consequences”. He points to techno-terrorists inspired by grey goo (2026-04-12 Future-Were-Building).

Form is chosen for the reader. Future Rising’s sixty short reflections were meant to be “long enough to help carve out a quiet space for reflection, but short enough that they don’t add to the noise” (FR p.19). Modem Futura is “intentionally authentic and conversational… minimal editing”. This style “eschews any sense of being ‘talked-at’” (2024-12-22).

He works against institutional incentives, and names them. - The research-teaching-service evaluation framework “actively discourages public-facing work” (2026-04-12 Stick-Figures). - In 2016 he proposed counting online casual-learning resources “toward academic tenure and promotion” (nnano.2016.167 p.735). - After chairing ASU’s university promotion and tenure committee, he describes the “crab bucket”: an official message of creative freedom, while committees ask “What is your h-index?” (2026-09-24; AI-edited).

Scrappy on purpose. - Risk Bites was “an experiment… that leant into my limitations” by “a full blown academic with limited time and even less talent” (2024-09-04). - The million-view nanotech video was “by any professional standard, terrible”, yet “I can tell a story and connect complex ideas to everyday experience” (2026-04-12 Stick-Figures). - He admits the “no talent” line is “a little tongue in cheek” (2024-09-04, fn 5). - The video was “whipped up for a school teacher” at the last minute (2024-09-04, fn 1), a point that matches the 2016 column’s account (nnano.2016.167 p.735).

Un-disciplinary. “I’ve no idea what my discipline is these days. I am 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). He describes his ASU work as combining “physicist mindset, understanding of risk, innovation around how we think differently about risk, and love of engaging with people from different disciplines” (TT p.2). He also says the connecting philosophy is “the hardest part of my work to surface, precisely because it doesn’t belong to any single discipline or output” (2026-04-12 Future-Were-Building).

5. His role as he sees it#

6. What is distinctive#

What follows is framed around the AI discussion, since that is where he places his work.

  1. Thirty years of the same gap, seen from inside. He has seen the gap between capability and capacity to navigate from the lab bench (HSE, NIOSH), from federal coordination (co-chair of the NNI’s Nanotechnology Environmental and Health Implications working group), from Congress, from the Wilson Center, from World Economic Forum councils, and from the classroom. - That gives him the authority to say the hard problems “are about human cognition, institutional design, attribution, trust, and meaning” (2026-04-12 What-Thirty-Years). - He also knows the institutional realities: bylines that hide who did the work, proposals that fail and still shape what follows (2026-04-08). - He contributed to the history he now draws on. His 2014 column cites his own co-authored nanotechnology chapter in the EEA’s Late Lessons from Early Warnings (2013) (nnano.2014.116, ref. 1).
  2. A risk scientist who knows the numbers well enough to know their limits. He is neither an outside critic of quantification nor a technocrat. He can benchmark nanotube exposures and cite NIOSH limits, and also say a signed waiver was about shame (RR p.195). That lets him move the AI conversation off both “probability × harm” and “vibes”, towards threats to value that are hard to count but real: dignity, identity, belonging, agency.
  3. A mindset that dissolves the binary. Risk as threat to value, a landscape, navigation and orphan risks together turn “accelerate versus prevent” into a question of pathways. “There will be no single silver bullet” (2026-04-12 What-Nanotechnology). Precaution and permissionless innovation both become questions of timing, trigger points and who takes part.
  4. Imagination as a rigorous tool for a world moving away from the conventional. Few risk scholars put haiku, science fiction, short stories, games and pizza seminars at the centre of method. He has done so since 2015, and argued for it on risk grounds: imagination, not cost, is often the main barrier to responsible innovation (nnano.2015.35 p.200). Conventional frames make “categorical errors” with AI (2026-09-24). The initiative’s founding values put this into practice: - “Obsessive curiosity”; - “Radical creativity”; - “Respectful inclusivity”; - “Grounded exuberance”; - “Catalytic serendipity” (FoBH homepage).
  5. The human side of AI, before large language models. “What it means to be human” drove his work before ChatGPT. Examples: - Future Rising’s three “base codes” and the move from extrinsic to intrinsic technologies (2024-01-01; 2026-04-12 Future-Were-Building); - the 2014 question of “responsibility in the face of such audacity” about 3D-printed artificial minds (nnano.2014.294 p.956); - the Ex Machina reading of manipulation as “far more worrisome than superintelligence” (2026-04-12 Future-Were-Building).

So his 2025–26 ideas are extensions of long-held questions, not new positions: constitutive resonance, honest non-signals, “Seemingly Conscious AI”, and the argument that waiting for consciousness “means waiting too long”. 6. Hands-on experimenter who reports failures. He writes about AI from inside its use: courses, dissertations, bots, AI-native books. He reports what failed (“I was wrong”; “Boy, was I so wrong”; “some of this could be completely wrong”). He applies the same humility to himself that he asks of the field. 7. A public voice that treats readers as smart and says it cannot see the future. He refuses to preach, polarise or fear-monger, and assumes “most people are reasonably smart” (TT p.5). He also admits “really, really bad days” with AI (2026-09-24). That combination is uncommon in an AI debate dominated by confident prediction. 8. He shows humility and does not just claim it. Examples: “I hold it provisionally” (2026-04-12 Future-Were-Building); “I found seven” papers on epistemic vigilance and AI (2026-04-12 Honest-Non-Signals); “probably incomplete”; “occasionally blinkered”. For something as poorly understood as AI, he offers calibrated uncertainty together with a willingness to act anyway.

7. The passages that best reveal how he thinks#

  1. RR pp.193–200, Rethinking Risk (2017). His whole method in one short essay. It questions a common term (“risk aversion”), uses fiction as evidence, tells a self-deprecating story of a physicist failing at his own numbers, and moves from risk as danger to risk as revealing value. It ends by thanking “‘yes and’ collaborations” for helping him see the world “through new eyes”.
  2. nnano.2015.196 pp.730–731, Why we need risk innovation (2015). The founding statement. It covers the landscape, threat to value, “parallel innovation” in risk, “risk entrepreneurship”, “serendipity”, and the seventeen haiku next to Tox21. It shows play and creativity were part of the concept from the start.
  3. 2024-03-17 undergraduate-playgrounds-not-playpens. Mostly for the postscript: “So much of how I explore new ideas… is grounded in play”. It shows the method in action (Google play, emailing Resnick, admitting bias), and it is where playgrounds and playpens come from.
  4. TT pp.2–3 (2023). Physics as “sheer delight”, science as “a love language”, being wrong as “fuel”, “un-disciplinary”, and bridge-building “fast”.
  5. nnano.2016.28 pp.211–212, Navigating the risk landscape (2016). The quantitative foundation (benchmarking, statistical definitions of materials) combined with value, humility (“probably incomplete”) and the rule “quick to question, and slow to respond”. It also covers acting on early warnings “even before the science is mature”.
  6. FR pp.148–151 and 166–167, Complexity, Hubris, Change. Against false precision and the illusion of control, while keeping the mathematics of change, “when tempered with humility”.
  7. 2026-04-12 What-Thirty-Years. His own map of the whole career: “The pattern hasn’t” changed; “the gap is never primarily technical”; AI as “a human question”.
  8. 2026-04-12 Stick-Figures. How he sees his role: the privilege-and-obligation argument, scrappy videos, films as “democratic entry points”, and the honest broker.
  9. 2024-12-22 why-modem-futura-is-more-than-just-another-tech-podcast. The initiative’s founding intent in his own words (“catalyze thinking at scale”, not a research centre), thinking in public, serendipity, and “not… preachy or polarizing or boring”.
  10. 2026-04-08 Before-the-Fourth-Industrial-Revolution. How he values his own contribution: conditions for possibility, failed experiments as formative, making messiness visible.
  11. Nature 475:31 (2011) and nnano.2014.196 p.658–659. Changing his mind in public, and then standing by the evidence against a fashionable alarm.
  12. 2026-09-24 being-an-academic-in-an-age-of-ai (AI-edited; the footnotes especially). His most recent account of the academic’s role. It argues that past frameworks make “categorical errors” with AI, and that speculation belongs “within a context of humility”.