Review: the “conventional box” in 05 and 06#

Working review, 27 September 2026. It checks 05-maynard-risk-and-ai-map.md (the map) and 06-huang-and-late-lessons-through-maynard.md (the report) against working/maynard/perspective/portrait.md and facets F1–F6. The question is where the two documents reduce Maynard’s thinking to conventional risk-governance categories (gates, evaluation, who decides, frameworks) at the expense of his mindset, and where that mindset is present but framed as an operational tool or as ornament.

Method. I read the portrait and facets first, then 05 and 06 in full. I did not open 01–04 or working/huang, working/late-lessons or working/synthesis; where 06 cites them, I report its citation. Key quotations below were re-checked against the text mirror (futureofbeinghuman) on 27 September 2026. The exceptions are book and paper quotations, which are taken from the portrait and facets, where they were already verified. Posts are cited by date and slug, papers by file or DOI stem and page, and Films from the Future as FFTF with its page. Two flags are carried over from the portrait. [AI-edited]: the 2026-09-24 King’s lecture body, which is his spoken ideas smoothed into prose by a model. [mixed]: the 2026-07-16 orphan-risks paper. Line numbers refer to the files as they stood on 27 September.

The finding in brief#

Both documents are careful, well sourced and accurate at the level of positions, but they present him as a risk-governance scholar with an unusually broad definition of harm: - His mindset becomes a “value layer” added to a risk-science base. - Navigation becomes either a verb or a list of instruments. - The risk landscape becomes a coverage table. - Imagination appears mainly as the thing plausibility has to discipline. - Play, creativity, curiosity and serendipity are marginal in 05 and entirely absent from 06.

The two documents also go further than leaving things out. Both actively correct his own mindset language as an overstatement (issue 2). And 06 turns his work into a menu of governance mechanisms, then judges it by what a regulator would need (issue 5).

Word counts show the imbalance. The counts are raw, and the portrait is about a quarter of the map’s length.

None of the following appears in either document, although each is central in the portrait: - “this lack of creativity and flexibility … increases the chances of things going wrong” (2016-01-11); - “not thinking creatively enough” (FFTF p.174); - the 2015 “culture of experimentation” “epitomized by serendipity”, or the seventeen haiku set beside Tox21 (nnano.2015.196 p.731); - “opens up new possibilities rather than closing down conversations” (2023-05-31); - “jolted out of our existing mental and procedural risk-ruts” (nnano.2015.286 p.1006); - “grounded in play” (2024-03-17); - “avoid it or flip it, and so get to the good” (2026-09-24 [AI-edited]); - “as soon as we start evaluating it within past frameworks, we make categorical errors” (2026-09-24 [AI-edited]).

Ranked issues#

1. The account that sets the lens leaves out the mindset half (both documents; root cause)#

Location. - 05 §1, “Two points Maynard has emphasised (September 2026)” (lines 35–42). The title block says the map “reflects two points he emphasised”, and later sections refer back to them. - 06 §1.5, “Two points…” (line 79), “reflected throughout”, and §3.1–3.2.

Problem. Both documents build their lens on two September 2026 points: his approaches build on past learning, and his sparing use of numbers for AI is deliberate. The substance of his September 2026 account is missing. When a technology fits no earlier type of risk, the whole mindset about risks, benefits and the path between them has to change. Risk innovation, the risk landscape, navigating rather than managing, and threat to value are offered as mental models that open possibilities, not as the operational way to do things. And play, creativity, curiosity and serendipity are integral to breaking out of stovepiped thinking. Because both documents are instructed to reflect “throughout” only the continuity half, every later section leans towards continuity (see issue 2).

Evidence from his work. The portrait makes both halves one insight: “It had two halves, and he kept both” (portrait §4; F3 §2). - The mindset half: “old wineskins of conventional risk thinking” (FFTF p.23); “no easy guidelines or rules of thumb”, only “ways of thinking” (FFTF p.39). - On AI: risks “sliced, diced, and solved, using a conventional mindset” (2023-05-31). - The mental-model intent: “designed to open up new ideas and possibilities” (2016-01-11). - The founding principles of the Future of Being Human initiative: “Obsessive Curiosity,” “Radical Creativity,” “Grounded exuberance,” and “Catalytic Serendipity” (2026-09-20 reasoning-llms-just-want-to-have-fun, n.2).

Fix. Add a third point to 05 §1 and to 06 §1.5, and have later sections refer back to it, as they now do for the other two. Suggested wording (analysis register, not his voice):

3. When a technology fits no earlier type of risk, the mindset has to change. Maynard has emphasised (September 2026) that risk innovation, the risk landscape, navigating rather than managing, and risk as a threat to value are mental models offered to open up possibilities, not procedures. Play, creativity, curiosity and serendipity, the founding principles of the Future of Being Human initiative, are integral to how he thinks conventional, stovepiped frames can be escaped. This is the other half of point 1: the foundations are kept, and the frame built on them changes.

Then state explicitly that points 1 and 3 are held together, and that neither should be read as qualifying the other away.

2. “Built on” has been overcorrected into “just a layer”, and his own mindset language is corrected as overstatement (both)#

Location. - 05, in five places: - the §4 preamble (line 253): “his positions are easily made to sound more absolute than they are … the substance is usually an extension”; - C3 (line 275): “Some of his own summaries sound more absolute than his practice, as when a 2026 essay recasts risk ‘from something to be minimized to something to be navigated creatively in pursuit of value’”; - T9 (line 703): “new ideas are added as layers on older ones”; - §7 “Layers, not replacement” (line 756); - “How he updates” (line 848): “the latest layer of a record”. - 06, in two places: - §3.1 “Building on, not replacing” (line 186): “His own shorthand can sound more absolute”, quoting the same line; - §11 (line 1036): “Dropping those hedges would present him as more absolute than he is”.

Problem. The map and the report treat “navigated creatively in pursuit of value” as rhetoric to be tamed. It is actually one of his most exact self-descriptions, and it matches his September 2026 account. Reading “built on” as incremental turns the change of mindset into an add-on. The portrait lists “Reducing risk innovation to an operational add-on, toolkit or compliance layer” as the first common way of getting him wrong (portrait §10). The layer metaphor also makes the new element sound optional. For him it is the necessary response to technologies that “run out of steam” the old approaches (FFTF pp.22–23).

Evidence. - “Without risk innovation, all we are left with is business as usual” (nnano.2015.196 p.731). - “we need to be jolted out of our existing mental and procedural risk-ruts” (nnano.2015.286 p.1006). - Regulatory frames are “usually not remotely the right shape, never mind being an adequate fit” (2016-01-11). - “as soon as we start evaluating it within past frameworks, we make categorical errors” (2026-09-24 [AI-edited]). - The continuity half is also his: “an evolution of the old black-and-white mathematics of risk” (RR p.200, 2017_Rethinking-Risk_VVEV-CSI-chapter.pdf). But “evolution” names a new frame grown from old foundations, not a layer on top.

Fix. - Replace “layer” language with “a new frame built on kept foundations” throughout both documents, and rename 06 §3.1 “A changed mindset on kept foundations”. - Delete the sentences that correct “navigated creatively”, or turn them round: quote the line as his summary and note that the “skeptical of both” sentence is how he holds it, not a retreat from it. - In 05 §4, change the caution to warn against both errors: making him sound like a revolutionary who discards risk science, and making him sound like an incrementalist who only adds items.

3. Play, creativity, curiosity and serendipity are missing as method and as a risk competence (both)#

Location. - 06: no section, bullet or lens mentions them as his method. Films appear only as a source of false alarm (“a narrative that belongs more in science fiction”, §4.2, line 365). - 05, in four places: - the architecture’s “Manner” line omits them (“public reasoning · humility · stories and films · self-implication · building to think”, lines 174, 221–222); - “His method, in brief” names four habits without them (lines 407–415); - T8 “How he thinks” gives the core as “an evidence conscience … humility … a discipline of plausibility” (line 697); - none of the eighteen commitments or the fourteen constants (lines 782–797) concerns them.

Problem. These are not his style. They are his argument about how risk is seen at all, made in the language of risk since 2015, and his account of how to escape “bounded infinities” of conventional thought. Leaving them out drops the part of his method that the portrait calls most distinctive for a world “diverging rapidly from the conventional” (F1 §9 item 8). It also breaks 05’s own weighting rule. An idea is “central when it spans many years, organises other ideas, and recurs unprompted” (05 §1, line 50). This one spans 2015 to 2026, organises the films, the playgrounds, the transitions models and risk innovation itself, and recurs unprompted.

Evidence. - 2015: risk innovation “encourages a culture of experimentation — a culture grounded in transdisciplinarity, creativity and imagination; and epitomized by serendipity”. His first worked example is “a book of seventeen haiku” (nnano.2015.196 p.731). - 2016: “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). - 2018: AI risks “may blindside us, in part because we’re not thinking creatively enough” (FFTF p.174). - 2021: “the juxtaposition of seemingly unrelated ideas can jolt us out of conventional ways of thinking” (2021-04-09 bounded-infinities-quantum-tunneling-and-the-future-of-education). - 2024: “So much of how I explore new ideas … and revel in the serendipity of new discoveries, is grounded in play” (2024-03-17 undergraduate-playgrounds-not-playpens). - 2025: the playpen “quickly falls apart where … the journey is breaking new ground” (2025-03-15 ai-playgrounds-in-higher-education). - 2026: “it would be embarrassing if we were all wiped out by something because we didn’t have the imagination to foresee it” (2026-09-15 will-ai-really-kill-us-all, n.5). “much of my work uses play, creativity, and serendipity, to explore new ideas in unexpected and often deeply insightful ways” (2026-09-20, n.1). “the joy of playing around and serendipitously discovering something” (2026-09-24 [AI-edited]). - What play has produced: the illusion of reciprocity (2023-04-05), stochastic agency (2024-10-27), and the transitions quadrant that came from “Experimenting with the ladder” (2024-08-18).

Fix. - 05: - Add a commitment next to C2 and C3: “Seeing risk takes imagination. Creativity, play, curiosity and serendipity are how thinking escapes frames that a technology has outrun, and plausibility and humility discipline them.” Give it a firmness line (from 2015, named as play from 2024, defended in 2026) and a source list. - Add concept entries: risk entrepreneurship and a culture of experimentation (2015); creativity as a risk competence; bounded infinities and metaphorical quantum tunnelling; designed serendipity; grounded exuberance. - Rewrite the core of T8 around the portrait’s cycle: something catches him, he questions the frame, loosens it, tightens it, builds or tests, publishes provisionally, and revises in public. - Add the thread to “What stays constant”. - 06: add a lens section before §3.1 (see issue 4).

4. 06 turns a way of seeing into positions scored for agreement, and the mindset divergence with Huang goes unnamed#

Location. - 06 §2 “What Maynard’s work brings” (lines 97–104): seven bullets, all positions. - §3: ten lens subsections, none on how he thinks. - §4–5: alignment and divergence. - The §5 divergences (lines 119–129) are all governance categories: scope of “safety”, who decides, incentives, tests, public action, tempo, form of confidence, risk talk and jobs. - §5.1 (lines 475–490) comes closest but treats the issue as a dispute over “what AI is”.

Problem. The portrait says he offers “a way of standing in front of a technology”, not “a new list of risks, a governance mechanism or a forecast” (portrait §8). F1 says “Method before verdict, so no camp”. Scoring his work point by point against Huang’s leaves out the most important contrast. Huang’s frame (“In the final analysis, engineers are doing engineering work… we understand it, obviously”, [1:10:03], as quoted in 06) is a live instance of the conventional mindset his work says fails for AI. The contrast between Huang’s “Enough predictions” and Maynard’s “some degree of informed speculation, and some degree of imagination” (2026-09-24, n.4) is a real divergence about how to know. In 06 it appears only as a limit under an alignment (§4.1, line 355).

Evidence. - “treating the new as if it’s something old” and “metaphors are never completely neutral” (2026-02-22 what-we-miss-when-we-talk-about-ai-harnesses). - “as soon as we start evaluating it within past frameworks, we make categorical errors” (2026-09-24 [AI-edited]). - In his series introduction he corrected his own first impulse to read Huang “within their own frame and agenda” (as quoted in 06 §1.6).

Fix. - Insert a section before §4, “The difference in mindset”, with three parts: - (a) Huang’s engineering frame, and the gate frame that 06 inherits from 03, as instances of the conventional mindset; - (b) what his mental models do with the same material: threat to value, landscape, navigation, and imagination as risk perception; - (c) the imagination divergence stated as a divergence. - Put these first in the §2 list of divergences. - Rewrite the §2 “What Maynard’s work brings” bullets to lead with the mindset and its imaginative means, then the positions.

5. 06 turns his work into a governance programme and judges it by regulatory yardsticks#

Location. - 06 §2 “Approaches it points to” (lines 139–151). - §7 (lines 700–763): fourteen items, including “in control” criteria, independent evaluation, evaluation in use, exposure measures, mandatory incident reporting, rules for every firm, permission scaled to reversibility, property-based triggers and one standard of evidence. They are assessed by “the tests 02 §10.2 and 03 §11.3 apply” (line 702). - “What his work does not supply” (lines 764–769): an evidentiary bar, an operational test, a threshold, evaluated tools and an evaluation method. - 05 §8 “Gaps” and tension 3 do the same, more lightly.

Problem. The measuring stick is the one his work sets out to change. The deficits are real, and the portrait names them (portrait §9, “Mindset over tool”). But they are presented without his reason for them, and without saying what the mindset supplies instead. In a domain nobody yet understands, a procedure would be false precision (F1 §9). The portrait also warns against reading him as an advocate for a policy menu: he is an honest broker for “process and capacity” (portrait §10, item 18).

Evidence. - “no easy guidelines or rules of thumb” (FFTF p.39). - “What the Planner does not do is provide answers to problems”, and “the Planner is a catalyst” (2023-11-21 ai-and-risk-innovation). - “No recommendations at this point – remember the humility bit” (2025-02-04 openai-deep-research-ai-scholarship, his prompt). - “I don’t have a governance solution for AI. I’m not sure anyone does” (2026-04-12, andrewmaynard.net, as cited in both documents).

Fix. - Restructure §7 in two parts. - (a) Moves in thinking his work brings to this debate: - question the frame (“just software”, “in control”, “harness”); - ask what is valued, by whom, and what is aspired to; - map the landscape, opportunities included; - look for where a threat can be “flipped” into an opening; - speculate with discipline “within a context of humility”; - convene unlikely voices; - watch the vocabulary. - (b) Where an instrument is needed, the present items, framed as what those moves might lead to. - Retitle “What his work does not supply” as “What it does not supply, by design, and what it supplies instead”, and pair each gap with his rationale. - Replace the 02 and 03 tests as the only yardstick, or add his own tests beside them: does it open possibilities rather than closing down conversations, and does it widen who can think well about the problem?

6. Imagination appears mainly as the thing plausibility disciplines (05)#

Location. - 05 C8, “Discipline imagination with plausibility” (lines 319–325). - T3: “2018: manipulation over superintelligence, plausibility over imagination” (line 669). - The crosswalk row “Epistemics of the future” (line 234). - Lens B2, “Is this plausible, or only imaginable?” (line 929).

Problem. Only the constraining half is a commitment. The generative half, imagination as how risks and possibilities come into view, shows up only as caveats (FR p.86; “informed speculation”). T3 inverts his 2018 argument. The manipulation judgement it summarises as “plausibility over imagination” came, in his words, from the failure to imagine: “we’re not thinking creatively enough about how an AI might threaten what’s important to us” (FFTF p.174).

Evidence. - “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” (FFTF p.282): both halves in one breath. - “the respect, due diligence, and intellectual rigor and imagination, that they deserve” (2023-05-31). - “some degree of informed speculation, and some degree of imagination” (2026-09-24, n.4).

Fix. - Retitle C8 “Imagination and plausibility together: creativity to see, plausibility to rank”, and open it with FFTF p.282 and p.174. - Change T3 to “manipulation over superintelligence, found by imagining how AI could threaten what people value, and ranked by plausibility”. - Pair B2 with a new lens (issue 10).

7. Navigation is reduced to a verb or a list of instruments (both)#

Location. - 05: there is no concept entry for navigation. T1 has “His operative verb is ‘navigate’, never ‘eliminate’” (line 630). T4 has “His response is to navigate, not prohibit: regulate designed manipulation, consider pausing…, place a duty of care…, put the safety message first” (line 673). Lens group E, “How can it be steered, and by whom?”, has the landing “builds portfolios rather than looking for silver bullets” (lines 955–964). - 06: “navigate” appears four times. §3.8 treats steering only as the opposite of stopping and of inevitability. §4.3 is titled with Klein’s “the flip”, meaning a reallocation of compute to verification. Maynard’s own “flip it” does not appear.

Problem. For him, navigation is a stance with a physics behind it, not a synonym for governance. It rests on chaos with bounds: no full control, but limits and points of leverage. It calls for a map rather than a forecast, steering rather than control, and course correction. Its aim is value, so a threat can sometimes be turned into an opening and not only reduced. Management stays, as the operational layer inside the stance (portrait §4).

Evidence. - “we cannot wield perfect control”, yet futures “can be squandered if we don’t think ahead” (FFTF p.41). - “Navigating the risk landscape” (nnano.2016.28). - “traditional ‘set it and forget it’ management doesn’t work” (2025-05-18 exploring-ai-through-cause-and-effect). - “remove risks, help identify ways to circumnavigate them, or strategically absorb them” (2023-11-21). - “you can only begin to realize the benefits … so that you can navigate around it … avoid it or flip it, and so get to the good” (2026-09-24 [AI-edited]). - 2024-08-18 four-ways-of-thinking-about-advanced-technology-transitions: a framework whose axes are “degrees of freedom” and “the mindset that transitions were approached with”.

Fix. - 05: add “Navigation rather than management” as a Core concept in §5.1 or §5.4 (chaos with bounds; map, not forecast; course correction; flip; management as the operational layer). Rewrite the T4 sentence so that the instruments are what navigation uses, not what it is. Retitle lens group E “How can a way through be found, and by whom?”, with a lens on flipping. - 06: set Maynard’s “flip it” beside Klein’s “flip” in §4.3 and show the difference. Klein’s reallocates effort to verification. Maynard’s turns a threat into an opening towards value. Then apply it to the skills exchange in §5.2: Huang’s “We’re going to discover new ones” becomes a navigation question (what to preserve, what new capacities to build, and how), not only a scope-of-safety dispute.

8. The risk landscape becomes scenery or a coverage table, and his own mindset model is barely used (both)#

Location. - 05 §5.1: “Risk landscape | The terrain between an idea and its realisation, to be mapped and navigated, not eliminated | … | Recurring” (line 440). His Avoid/Adapt/Extend/Embrace model gets one row in §5.4 (line 514). - 06 §8.6 (lines 837–856) converts his reservation about “a broader landscape of emergent AI characteristics, capabilities, threats, risks, and benefits” into a table of which threat and opportunity cells 01–04 covered, using the 2024-08-25 quadrant. §9.1 (line 864) reads the moment as an “early-window moment” with two questions: who decides, and can sincere people see enough.

Problem. The landscape is a mental model of terrain that technologies “both face and help to form” (2016-01-11). It is full of people, power, shifting ground, tipping points and stable points. Reducing it to cells to be ticked is operational. The 2024-08-18 quadrant, his own model with “mindset” as an explicit axis and precaution as one legitimate posture of four, is the natural tool for 06. It is absent from 06.

Evidence. - “It opens up new ways of imagining the risk landscapes that new technologies both face and help to form” (2016-01-11). - “shifting hills and valleys” (nnano.2016.28 p.211). - The quadrant’s axes are “a tendency to try and maintain things as they [are] … versus a willingness to embrace change”, and “With a creative mindset … it becomes possible to push potentially disruptive tipping points far into the future”. It is offered as something that may belong in “the trash can of bad ideas” (2024-08-18).

Fix. - 05: promote the risk landscape to Core, with a line on what it opens up. - 06 §8.6: keep the coverage table, but first read the reservation as a failure to see the terrain. Then use the 2024-08-18 quadrant to place the actors: Huang’s “accelerate” and “discover new ones” in extend or embrace, the Late Lessons reports mainly in avoid, and the labs’ containment and frameworks in avoid and adapt. Say what each mindset makes visible and what it forecloses. - Add a third question to §9.1: are the frames in use (containment, release, “just software”, “in control”) the right shape for what happened, and what would it take to imagine what they miss?

9. Threat to value is presented as an extra definition and a list of harms, not as a way of seeing what matters (both)#

Location. - 05 §2 “The engine” (line 113) and C2 (lines 263–269): “introduced on top of the probability-of-harm definition”. - 06 §3.1 “The value layer” (line 184). - 06 §7 item 1, “add a value layer; widen what ‘safe’ covers” (line 704), and item 7, “reach builders through what they value”.

Problem. The frame appears mainly as a wider harm list to bolt onto safety frameworks, or as a lever on firms. The portrait’s reading is stronger. It opens up aspiration and dignity. It turns go/no-go choices into design questions. It makes resistance intelligible. It puts forgone benefits in the same account as harms. And it turns risk into a way of seeing what matters. 05 does credit part of this (C1 and C10 on thriving and on concern as a signal). 06 uses almost none of it.

Evidence. - “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). - Conversations “can be elevated from simplistic ‘go/no-go’ options”, which “opens the door to creative and innovative approaches” (RR pp.197–198). - “something we aspire to and cannot bear to lose sight of” (FFTF p.24).

Fix. - In both documents, lead the threat-to-value entry with what it opens, and move the “extension of the probability definition” point second. - In 06, apply it as a design question to the interview itself. What does each party value? Huang values ambition, purpose and “benefits them”. Klein values skills, democratic accountability and who carries the worry. What design choices would protect or grow both? At present the frame appears only as a scope extension to gates.

10. The lenses in 05 are a governance checklist with no lens for the mindset#

Location. 05 §9 (lines 912–973). There are twenty-nine lenses in six groups. “Who decides” (D) has seven lenses and “steering” (E) six. The only frame lens, F4 “What does the framing make invisible?”, comes last. Each group opens with “His usual landing”.

Problem. Anyone using §9 on new material, as 06 did, gets a governance audit. There is no lens that asks whether the frame fits, what is failing to be imagined, what juxtaposition or story might jolt the analysis, whether a threat can be flipped, or which mindset is in play. “Landing” also suggests a verdict, where the portrait describes him as aiming at “a map of pathways, not a ruling” (portrait §2).

Evidence. As in issues 3, 6, 7 and 8. Also: “a framing of AI risks and benefits that opens up new possibilities rather than closing down conversations” (2023-05-31).

Fix. Add a first group, “A0. Is the way of thinking fit for this?”, and fold F4 into it. Suggested lenses: - Is the technology being squeezed into a frame built for something else (FFTF p.23; 2016-01-11)? - What are we failing to imagine about how it could threaten, or create, what people value (FFTF p.174)? - What unlikely source, juxtaposition or story could jolt the analysis (2021-04-09; 2026-09-24, n.8)? - Can the threat be avoided, circumnavigated, absorbed or flipped (2023-11-21; 2026-09-24)? - Which mindset (avoid, adapt, extend or embrace) is each actor bringing, and what does it foreclose (2024-08-18)? - Is curiosity, joy or wonder itself at stake (2024-11-10)?

Change “His usual landing” to “Where he usually heads”.

11. The architecture puts risk science at the centre and method at the edge (05)#

Location. 05 §3 “The architecture” (lines 164–174) and the diagram (lines 196–223). The engine is “value … on a base of risk science”. Imagination sits under “Disciplines” as the thing to be disciplined. “MANNER, throughout” is a footer.

Problem. The picture reproduces the box. The change of mindset, the stance of the whole enterprise, has no box, and the imaginative method that drives it is a footnote of manner. The portrait’s structure is different: a question about people drives a mindset for technologies that do not fit. That mindset is built on risk science, loosened by imagination and tightened by plausibility.

Fix. Redraw the diagram. Put a “Stance” box (a changed mindset for technologies that fit no earlier risk type) between Purpose and Engine. Add a “Method” box (curiosity, play, story, juxtaposition, building to think, loosened then tightened) with an arrow into the Engine labelled “how risks and possibilities come into view”. Keep risk science as the base of the Engine. Add “play · curiosity · serendipity” to the method line.

12. 06 reads his Late Lessons reservation as a content gap, not a critique of the thinking mode#

Location. 06 §8.5–8.6 (lines 814–856). His remark that the model “struggled to apply conceptual rather than literal comparisons … (for instance claiming that AI is not biology …)” is read as a lack of receptor-side harms and as missing quadrant cells, with the note that “he may have meant something more literal” (line 833). 06 §3.3 turns transfer into a three-mode taxonomy (literal, conceptual, principle).

Problem. His complaint is about category-bound thinking, the very thing his method exists to escape. For him, analogy is a creative probe whose break points produce insight. It is not a classification exercise. The fix is not only to add items but to run the analogy.

Evidence. - “Of course, an algorithm is not a chemical”, yet he found the analogy “intriguingly compelling” and built algorithmic exposure from it (2019-03-05 should-we-be-treating-algorithms-the-same-way-we-treat-hazardous-chemicals). - The juxtaposition method (2021-04-09). - His series introduction warns against “shallowly interpreting Huang’s comments within their own frame” (as quoted in 06).

Fix. Add a paragraph to §8.5. It should say that the reservation concerns a mode of thinking, and that his remedy is to carry the structure of toxicology (agent, receptor, dose, population, latency) across as a probe and treat each break as information. 06 already does this well once, with leaded petrol as a population-scale cognitive harm (§10.2). Say that this is his method, not only a transfer category.

13. 06 turns his way of engaging opponents into a rebuttal template, and treats his curiosity about Huang as a limitation#

Location. 06 §1.6 (line 91) and §6.5 (lines 678–696). His corrected first impulse and his wish to “understand what is a far more nuanced landscape” appear only as the reason every claim is Implied or Inferred. The engagement pattern is summarised as “Grant what is right; extend fairness to the person; bring the history …; and end on who decides”.

Problem. The template leaves out his opening move, curiosity about what drives the other person. It also leaves out questioning the frame and looking for what the other frame opens, and it makes “who decides” the fixed endpoint. The report never asks what in Huang’s outlook his work would find generative. It has a clear candidate in shared exuberance about technology, which is grounded differently in each man. And it has a direct film link. The inventor in The Man in the White Suit lacks “social curiosity”, the curiosity “to ask people what they think, and what they want” (FFTF p.222). That is a mindset remedy for Huang’s “that’s my problem” (§5.3), not a gate remedy.

Evidence. - He invited Jim Thomas to write on his blog because he “wanted to get a better understanding of how they saw the emerging relationship between society and innovation” (FFTF p.191). - “I revel in their potential” (2023-10-19 marc-andreessen-ditch-sustainability). - “grounded exuberance” (2026-09-20, n.2).

Fix. - Recast §1.6 so that his self-correction becomes the report’s method: curiosity first, the frame questioned, and the other side’s frame explored for what it opens. - Add these steps to the §6.5 pattern and drop “end on who decides” as the fixed endpoint. - Add a short passage in §6.1 on shared exuberance and the difference in how each grounds it.

14. Films and stories are used as sources of argument, not as instruments (both)#

Location. In 06, fiction appears only as a cause of false alarm (§4.2). The FFTF films are mined for sentences, such as Stratton as “myopically benevolent science” in §3.7, but no story is used as a lens. In 05, “Method in brief” treats stories as “instruments of foresight” and stresses that they “came relatively late” (line 412), without tying them to value-based risk.

Problem. Stories are central to his method because drama is built from threatened value. They are the natural instrument of the threat-to-value frame, not decoration. The portrait counts “Treating play, films and humour as ornament” among the ways of getting him wrong.

Evidence. - “Each of these films has a risk-based narrative tension that keeps its audience hooked” (FFTF p.23). - Hammond’s dream, Tommy’s hope and Kusanagi’s sense of self (FFTF p.24). - “Preach to someone about the future, and most people will shut down” (2024-01-21 how-can-stories-unlock-pathways-to).

Fix. - 05: add one sentence linking stories to threat to value, and move “came relatively late” into the history of the method rather than its description. - 06: use at least the two film readings his own texts supply for this material, Stratton’s lack of social curiosity (issue 13) and Her on designed intimacy (2024-05-21 openais-problem-with-the-movie-her), as lenses on the interview’s blind spots.

15. Where play appears, it is filed under education, values or tensions (05)#

Location. - 05 §5.10: “Play, serendipity and learning by not trying to learn | Bounded infinities and metaphorical tunnelling” (line 607), under education. - §5.9: “Joy, wonder and play” (line 594), under values. - Lens E4: play as how people “live with it intentionally” (line 962). - Tension 7: “argues for ‘permission to play’ (2025-03-15)” as evidence of the adopter-versus-critic tension (line 868).

Problem. Bounded infinities is his theory of how thinking escapes frames, not a teaching tip. Listing “permission to play” as a symptom of his entanglement recasts his method as a liability. The portrait resolves the apparent tension. Playgrounds have rules (“be kind, don’t spoil things for others”, 2025-03-15). Context and reversibility separate play from permissionless innovation (“I’d put breaking people, governance, society, and the planet, in this category!”, 2025-03-02 the-lure-of-permissionless-innovation, n.2).

Fix. Move bounded infinities and tunnelling into §5.5 (epistemics). Keep a pedagogy entry in §5.10 that points back to it. Remove “permission to play” from tension 7, or add the reconciliation in his own terms.

16. The evolution narrative leaves out how the method developed (05 §7)#

Location. 05 §7: the phases table (lines 725–735), “What stays constant” (lines 782–797) and “How he updates” (line 848).

Problem. The table traces risk concepts and AI concerns but not the method. It misses creativity written into risk innovation in 2015, bounded infinities and designed serendipity in 2021, play named as the root of his method in 2024, and joy defended and play pre-registered in 2026. It also misses analogy moving from confident translation to probe, and his move from observer to participant-instrument (portrait §2; F1 §6).

Fix. Add a “method” column or bullets to the phases table and a constant for “imagination as part of risk thinking (2015–2026)”. Add to “How he updates” that hands-on play is one of the main ways he changes his mind (2023-01-31, 2024-10-27, 2026-02-08 beeswax-hallucinations-and-ai-inventions).

17. Orphan risks are called “a tool” and an “operational edge” (both; lower priority)#

Location. - 05 T1: “one tool in this kit, the operational edge of the threat-to-value frame” (line 634). - 05 §5.1: “Recurring (a tool within risk innovation)” (line 453). - 06 §3.9: “one tool among several” (line 321); §11 “one tool here, not the centre” (line 1050).

Problem. The proportion caution is sound. Both documents rightly resist over-weighting a 2026 paper of mixed provenance. But “tool” misdescribes a way of seeing institutional blind spots. The portrait calls it the concept that changed most, and possibly his most valuable framing for AI now (portrait §4, §8).

Evidence. Orphan risks are “hard to quantify threats to value that often slip between the cracks of conventional risk approaches” (2020-10-15 the-ethics-of-advanced-brain-machine-interfaces-and-why-they-matter). The question he poses is “by what process does a known risk come to be nobody’s responsibility?” (2026-07-16 [mixed]).

Fix. Call it “a recurring lens” or “a mental model within risk innovation” and keep the proportion note. Reserve “tool” for the Planner and the register.

18. 06 imports the gate vocabulary of 03 and speaks for his work in it#

Location. 06 uses “gate(s)”, “gated” and “gating” seventeen times, for example “builder-held gates” (line 135), “who holds the gate” (§2, §5.3) and “release gates” (§7 item 1). The portrait and all six facets use the word once between them, in F3’s “evidence-gated action”.

Problem. When 06 states what his work “would” say in 03’s vocabulary, the question becomes who controls a checkpoint. His own questions are what is at stake, for whom, and which way to steer. Item 2 of §7 gets part of the way there (“Before who holds the gate comes who decides what it guards against”), but it stays inside the gate picture.

Fix. Keep “gate” when describing Huang, the labs or 03. When paraphrasing his work, use his vocabulary: value at stake, landscape, navigation, orphaned, flip, and “who decides what ‘safe’ means”.

What already works, to build on#

• 05. C1 (thriving as the point) and C10 (concern as a signal of value) carry the spirit of threat to value. §5.5 “Analogy as mindset, not template” and “Informed speculation with humility” are close to the portrait. Lens F4 (“What does the framing make invisible?”) is the seed of the missing lens group. “Stories and imagination as instruments of foresight” in the method note is the right instinct and needs linking to value.
• 06. §3.3’s “Errors in both directions” includes “not remotely the right shape” (line 223). It needs to be promoted from a transfer point to the mindset claim. §5.1 carries his reservation that the analysis treats AI “largely as it is depicted by Huang”. §6.5 records his fairness to opponents. §7 items 10 (vocabulary) and 14 (mapping pathways) and §9.6 (“humility is broader than adversarial thinking”) are the parts of §7 closest to his mindset, and could lead a restructured §7. §10.2’s leaded-petrol transfer is a good worked example of his analogy method.

One sentence both documents could carry near the top#

His risk concepts are offered less as procedures than as ways of changing how people think about technologies that fit no earlier category. They are built on quantitative risk science and held with humility against false precision. He argues that creativity, play, curiosity and serendipity are how that change happens, because a risk no one has imagined is a risk no one will see.