Revision log: 05-maynard-risk-and-ai-map.md, 27 September 2026#

What was changed in the public map of Maynard’s thinking, and why. Inputs: perspective/portrait.md, perspective/review/box.md, fixation.md and value.md, and facet F3 for detail. The pre-revision file is saved as perspective/review/05-map-before-revision-2026-09-27.md. Nothing in 01–04, 06, working/huang, working/late-lessons, working/synthesis or working/maynard-lens was opened.

Length went from about 40,200 to 54,100 words. Most of the increase is the new §2 (about 7,600 words).

1. Structural changes (read before revising 06)#

Sections renumbered. A new §2 was inserted, so old §2–§9 are now §3–§10:

Commitment IDs C1–C18 are unchanged. New content was merged into existing commitments instead of renumbering. Retitled: C3 (a changed risk mindset on quantitative foundations), C5 (humility as a working discipline), C7 (navigate rather than stop or simply manage), C8 (imagination and plausibility together), C11 (adds “known risks can become nobody’s“).

Lens IDs. - New first group M. Is the way of thinking fit for this? (M1–M7). - New C4 (does it act on the faculties used to judge and steer it?). - New D8 (how do risks become nobody’s?). - F4 removed; its question is now M2. - Groups D and E retitled. “His usual landing” is now “Where he usually heads” throughout. - Lens count: 29 → 37.

Tension IDs. 1–15 keep their numbers. Tension 5 is recast. Tension 16, “Symmetry against ‘safety message first’”, was dropped and replaced by a new 16, “Mindset over tool”. Tension 7 is amended.

06 cites 05 by section, commitment and lens. Those citations will need updating against the tables above.

2. Required change 1: a front section on how he thinks and works#

Added §2, How he thinks and works, drawn from the portrait and re-verified against the sources. Subsections: - 2.1 the core (the haiku beside Tox21, delight in physics, “very un-disciplinary”, the driving question, grounded exuberance); - 2.2 a changed mindset built on quantitative risk science (where it came from; the two halves; built on, not discarded; neither a revolution nor an add-on; why AI makes it unavoidable); - 2.3 the mental models and what each opens; - 2.4 imagination, play, curiosity and serendipity; - 2.5 the method cycle and humility as a working discipline; - 2.6 what matters to him; - 2.7 scholarship in public; - 2.8 the public scholar; - 2.9 what this brings to AI, with an assessment of each framing’s strengths and weaknesses; - 2.10 reading the rest of the map, with common misreadings.

Supporting changes. - Header and §1 “What it is for”: the map now says it sets out how he thinks before what he concludes, and why. - §3 opens by saying it should be read through §2. Its method-and-voice paragraph now points to §2. - §4: the architecture gains a Stance part and a Method part, and “Manner” becomes “Voice and role”. The diagram is redrawn so that the method feeds the engine (“how risks and possibilities come into view”) and risk science sits as its foundations. - Crosswalk: new first row, “How he thinks: mindset and method”. The stale §6 numbers in the table were corrected. - “His method, in brief” after §5 is rewritten around the method cycle. - T8 is rewritten as the cycle, and its development line now includes the method’s history.

3. Required change 2: the framings as mental models on quantitative foundations#

Layer language removed. The stack metaphor (“base layer”, “second layer”, “value layer”, “Layers, not replacement”, “latest layer”) is replaced by a foundation-and-frame account. The foundations are kept; the frame changes the questions the tools serve.

Places changed: - §3 engine; - §4 architecture; - C3 (retitled, “second layer” deleted); - C4 firmness line; - T9; - §8 formative-layer findings, “What stays constant” item 1, and “How he updates”; - the quantitative-foundation row in §6.1.

“Sounds more absolute” removed. The two corrections of his own phrasing were deleted: the §5 preamble caution and the C3 sentence on “navigated creatively in pursuit of value”. That line is now quoted in §2.2 and C3 as an exact statement of his thesis, with the “skeptical of both” sentence as how he holds it. The §5 caution now guards against both misreadings, rejection of risk science and a bolt-on module.

Threat to value now leads with what it opens. In §2.3, §3, C2 and the §6.1 row it opens with what the frame makes possible: resistance made intelligible, go/no-go turned into design, benefits in the same account, reciprocity, and risk as “an inevitability that reveals what the primary value is”. The “extension of the probability definition” point comes second.

Risk landscape. Raised to Core, with what it opens (a map of pathways) and the four-ways model with mindset as an axis.

Navigation. - New Core concept row, “Navigating rather than managing”. - C7 rewritten around navigation: map, fixed points, course correction, circumnavigate or absorb, “flip it”. - T1: “navigate, never eliminate” becomes navigate rather than eliminate or manage as an end. - T4: the instruments are now what navigation uses, not what it is. - Lens group E retitled “How can a way through be found, and by whom?”. Lenses M4 and M5 added.

Orphan risks, weighted by value rather than frequency. - §1 Method: new paragraph, “Centrality in the record is not the same as importance for AI”. - §2.3: full account, with roots (emergent risk 2011, Libby 2011, “mundane risks” 2014) and the Late Lessons link through his 2015 “ignored or overlooked” (2018-12-15). - §2.9: assessed as possibly his most valuable framing for AI, with its limits. - C11 retitled and extended. C2 points to it. - §6.1 row: “Recurring in the record; Core as a mental model. Value for AI: high”. - T1: “one tool … operational edge” and “about 16 posts” removed. - T2: Late Lessons link added. - §8 retrospective: “one tool within risk innovation” removed. - New lens D8, “How do risks become nobody’s?”. B3 keeps the 2018 dismissal question and points to D8. - New connection in §4: orphan risks → late lessons.

Tools as catalysts. §2.3 describes the tools as small and aimed at mindset. The new tension 16, “Mindset over tool”, keeps the honest edge: between 2017 and 2020 the ideas were also offered as tools and a business case.

Navigator argument. Added to C15, T4, §2.2, §2.9, the §4 connections (new) and lens C4.

4. Required change 3: play, creativity, curiosity and serendipity as integral#

C8 rewritten. “Discipline imagination with plausibility” becomes “Imagination and plausibility together”. The generative half is now documented on risk grounds from 2015 to 2026 (NN 2015-03, NN 2015-09, 2016-01-11, FFTF p.174, 2021-04-09, 2025-03-15, 2026-09-15 n.5). Plausibility ranks what imagination finds.

T3. “plausibility over imagination” becomes “found by imagining … and ranked by plausibility”. The same phrase in the architecture now reads “imagination disciplined by plausibility”.

New §6.5 concept rows. - creativity as a risk competence (Core); - questioning the frame (Core); - bounded infinities and metaphorical quantum tunnelling; - play as method, playgrounds not playpens (Core); - designed serendipity; - grounded exuberance (Core); - building to think, the self as instrument (Core).

The “Core” key was changed from “under thirty” to “about a quarter of the glossary”, now 36 of 148 rows.

Related concept changes. - “Analogy as mindset” is now “Analogy as probe, not template”. - The science-fiction row now ties stories to threatened value. - “Joy, wonder and play” in §6.9 is marked as values at stake and points to §6.5. - “Play, serendipity and learning by not trying to learn” in §6.10 is now a pedagogy entry that points back to method. Bounded infinities is no longer filed under education. - New §6.11 row: questions, not conclusions; convening. The humour row is expanded.

Stories. Linked to threat to value in §2.4, C8, the method note and a new §4 connection. “Came relatively late” is replaced by a dated history: art and speculative design in 2011–12, films in 2018. Play is placed at the root of his physics.

Tension 7. “Permission to play” removed as evidence of the tension, with his reconciliation added (rules, reversibility). An interpretive note on access was added (the portrait’s “Access, reach” edge).

§8. - New subsection, “How the method developed”. - New constants 15, imagination as part of risk thinking, and 16, navigation as the stance. - “How he updates” now includes hands-on play and experiment.

§9 and §10. - Open questions gain “Is the frame the right shape?”. - Lens group M comes first, and the §10 intro says the lenses open questions rather than forming a checklist.

5. Required change 4: fixation and evergreen fixes#

Private remark removed. The “Two points Maynard has emphasised” section and every “(§1)” pointer to it were removed. So was the phrase “the hubris of risk assessment”, which appeared about 10 times: in the header, §3, C5, the §6.1 row, §7 T1, §8 and lens B6. - It is replaced by a short, labelled personal-communication note that gives his whole September 2026 account (mindset, mental models, play, foundations, humility). It is not used as evidence, and each point is grounded in published work in §2. - The concept is renamed “Humility about precision: numbers that comfort without informing”. It is anchored on 2023-11-26 for AI and on PEN 2006, 2020science 2009 and FR p.148 for the long record.

Inevitability. - “The boat has already left the harbor” is dropped as evidence everywhere: §3, C7, the §6.4 row, T6, §8 “What changes” 5 and tension 5. - “We can’t pause it” appears once, as the working assumption he flagged as possibly flawed. - The position is now stated from NN 2015-12, FFTF p.288, 2025-03-30, 2025-08-31, 2024-10-27 and Dune 2024. - “then treated as unavailable for AI in general (2026)” is deleted. - Tension 5 is recast as “An inevitable trajectory with a shape still to be chosen”.

April 2026 essays. - The governance line is now quoted with its following sentence, in §3 and C13. - The early window is anchored on Testimony 2008 p.7, NN 2015-03 p.199 and NN 2016-03 p.212 (§3, C7, the §6.4 row, T6). - Convergence is anchored on 2015-01-30 and NN 2015-12 before FWB 2026.

2026-05-21 (routes through the excluded co-written book). - The C1 and C17 “In his words” quotes are replaced: C1 with 2025-03-30 and 2026-07-10, C17 with 2024-01-01. - The three-domain model is now cited to 2025-01-07 in §3, C17, the §6.4 row, T11 and Appendix A. - 2026-05-21 is kept only for his encyclical reading, “just a tool” and “cognitive surrender”. - It was removed from the C1 sources, the flourishing row and the tension 15 citation. - The book itself is not mentioned anywhere in 05.

Single sources re-weighted. - “Safety message first” is now a one-off 2026 wording of a long-standing stance: C10 (with FFTF p.227 and the 2018 Risk Bites recollection) and the §6.1 row, which becomes “Risk communication without alarm”. - Tension 16 on it is dropped. - “Exposure of the mind” moves from Occasional to Recurring (2019–2026). - The discontinuity claims are redated to 2023–26 (2023-04-12, 2024-05-05, 2025-03-15) in C14, T2, the crosswalk and §3. - The relational thread is described as running from 2023 in C15.

Evergreen wording. - Process statistics were removed from the body: 279/150 moved to Method; the 28% and 59% education shares, the “107 relevant posts” and “about 16 posts” are gone. - Relative time was replaced: “newest thinking” becomes “2026 thinking”; “Rising … prominent now” becomes “Rising (as of September 2026)”; “the concern he now calls” is gone; “where AI now sits” becomes “in 2026”. - Model names are now in the provenance notes only, with a one-line gloss (developer and year). The body says “AI-assisted” or “AI-drafted” except inside his quotations (“suckered by Claude”). This covered T1, T5, C11, the §4 connections, §6 rows, §8 “What changes” and tension 13.

Kept, as sound. The provenance rules and [mixed] flags, the formative layer (2005–2016), the changes-of-mind tables, the independent limits (whose value, the evidentiary bar, instrument and object), the governance history, and the whole AI-on-the-mind thread.

6. Verification#

• All new quotations in §2 and in revised passages were checked against the post corpus (working/maynard/corpus/), the FFTF text, the column and paper PDFs (Resources/maynard-papers/), and the April 2026 essays.
• Future Rising quotations (“as big and inclusive as possible”, FR pp.191–192) rest on the verified perspective notes (SUPP-self), because FR is not in the corpus.
• Corrections made while checking:
• “Lego and clothes pegs” is now “Lego and craft supplies”.
• “The privilege …” is lower-cased to match the source.
• The 2023-05-15 title is quoted in lower case, in §2.8 and T5.
• “Who is reaping …” is capitalised as in Bulletin 2008.
• “another commentary on the ethical challenges” (2019-11-01) is paraphrased, because a link splits the phrase in the source.
• 2024-04-07 is cited for his seminar’s community norms, and 2026-09-20 n.2 for the initiative’s guiding principles, instead of treating both as initiative values.

7. Open points#

• 06 needs its references to 05 updated (§1 above). It also needs the parallel fixes the reviews list for it; 06 was not touched.
• Evaluation. No evaluation of the risk-innovation tools exists. The map says so in §2.9, the gaps, and tension 16.