Bibliography
61 sources. Each entry gives what we use it for and where it falls short. Numbers match the citations on every page.
Classical or historical primary text
- [1]
Plato (trans. Benjamin Jowett) (c. 375 BCE). The Republic. Project Gutenberg edition.
- Findings
- Book VII sets out a higher curriculum of arithmetic, plane and solid geometry, astronomy and harmonics as preparation for dialectic: the mathematical arts train the mind to turn from appearances towards understanding.
- Limitations
- An idealised programme for a ruling class in an imagined city, not a description of actual Athenian schooling.
- [2]
Aristotle (trans. Benjamin Jowett) (c. 350 BCE). Politics. Project Gutenberg edition.
- Findings
- Book VIII argues that education should be public and should prepare citizens for the noble use of leisure (scholē), not only for useful occupations; it discusses letters, gymnastics, music and drawing.
- Limitations
- Assumes a society in which slaves and others perform necessary labour; its "free person" excludes most of the population.
- [3]
Aristotle (trans. W. D. Ross) (c. 340 BCE). Nicomachean Ethics. Internet Classics Archive (MIT).
- Findings
- Defines eudaimonia as activity of the soul in accordance with virtue over a complete life, and distinguishes theoretical wisdom (sophia), practical wisdom (phronēsis) and craft knowledge (technē).
- Limitations
- A virtue ethics rooted in the Greek polis; its account of who can flourish is restrictive by modern standards.
- [4]
Seneca (trans. Richard M. Gummere) (c. 65 CE). Moral Letters to Lucilius, Letter 88: On Liberal and Vocational Studies. Wikisource.
- Findings
- Questions whether the liberal arts make people good; argues they only prepare the mind for philosophy, the study of virtue. An early internal critique of the curriculum.
- Limitations
- A Stoic moralist's polemic, not a survey of Roman education.
- [8]
Hugh of St Victor (trans. Jerome Taylor) (c. 1120s–1130s (trans. 1961)). The Didascalicon of Hugh of St. Victor: A Medieval Guide to the Arts. Columbia University Press (Internet Archive scan).
- Findings
- Classifies knowledge into theoretical, practical, mechanical and logical, and names seven mechanical arts (fabric-making, armament, commerce, agriculture, hunting, medicine and theatrics) beside the seven liberal arts.
- Limitations
- Written for a monastic school; ranks the mechanical arts below the liberal ones.
- [13]
Craig W. Kallendorf (ed. and trans.) (2002). Humanist Educational Treatises. Harvard University Press (I Tatti Renaissance Library).
- Findings
- Texts by Vergerio, Bruni, Piccolomini and Guarino on the studia humanitatis: grammar, rhetoric, poetry, history and moral philosophy, combined with physical exercise and formation of character.
- Limitations
- Programmes written for elite boys (and, in Bruni's case, a noblewoman); not mass schooling.
- [14]
Jacob Burckhardt (trans. S. G. C. Middlemore) (1860). The Civilisation of the Renaissance in Italy. Project Gutenberg edition.
- Findings
- Popularised the image of the Renaissance "many-sided" or universal man (uomo universale), with Leon Battista Alberti as the exemplar and Leonardo as its culmination.
- Limitations
- A 19th-century interpretation; historians now regard its sharp contrast between "medieval" and "Renaissance" individuality as overstated.
- [16]
John Amos Comenius (trans. M. W. Keatinge) (1657 (trans. 1896)). The Great Didactic. Adam and Charles Black (Internet Archive scan).
- Findings
- Argues for teaching "all things to all people", girls and boys, rich and poor, in graded schools: an early programme of universal education and of pansophia, the unity of all knowledge.
- Limitations
- Grounded in a 17th-century Protestant theology of knowledge.
- [17]
Denis Diderot and Jean le Rond d'Alembert (1751). Système figuré des connaissances humaines (Encyclopédie, vol. 1). ARTFL Encyclopédie Project, University of Chicago (front matter).
- Findings
- The Encyclopédie's "figurative system" of human knowledge, adapted from Francis Bacon, divides all knowledge by faculty: memory (history), reason (philosophy) and imagination (poetry). A precedent for drawing knowledge as a map.
- Limitations
- Its faculty-based division reflects 18th-century psychology and places theology and the mechanical arts in contested positions.
- [18]
John Henry Newman (1852/1873). The Idea of a University. Project Gutenberg edition.
- Findings
- Defends liberal knowledge as its own end and the cultivation of an "enlarged" mind able to see the relations between fields; argues the university should not be reduced to professional training.
- Limitations
- A Catholic university apologetic of its time; dismissive of research as the university's core task.
- [19]
John Dewey (1916). Democracy and Education. Project Gutenberg edition.
- Findings
- Education as the continuous reconstruction of experience; learning through shared activity and inquiry; rejects the split between "liberal" and "vocational" education as a legacy of class division.
- Limitations
- Its optimism about schooling as a lever of democratic reform has been debated; implementation of "progressive education" varied widely.
- [20]
Franklin Bobbitt (1918). The Curriculum. Houghton Mifflin (Internet Archive scan).
- Findings
- Applies scientific management to schooling: derive objectives from the activities of adult occupational and civic life. A founding text of the "social efficiency" view of curriculum.
- Limitations
- Primary source for a view, not evidence that schooling became purely vocational.
- [32]
Kautilya (trans. R. Shamasastry) (c. 3rd c. BCE–3rd c. CE (trans. 1915)). Arthashastra. Government Press, Bangalore (Internet Archive scan).
- Findings
- Book I lists four fields of learning: philosophy and reasoning (anvikshiki), the Vedas (trayi), economics (varta) and government (dandaniti), an ancient classification that already includes economics and governance.
- Limitations
- Dating and authorship are debated; a treatise for rulers, not a general theory of education.
- [52]
John Maynard Keynes (1930). Economic Possibilities for Our Grandchildren. Essays in Persuasion (Economics Network archive).
- Findings
- Predicts that within a century the "economic problem" could be solved and the real problem would become how to use freedom from pressing economic cares; names "technological unemployment".
- Limitations
- Its 15-hour week did not arrive: productivity gains were taken mostly as consumption, not leisure.
Scholarly or general book
- [7]
William Harris Stahl, Richard Johnson and E. L. Burge (1971–1977). Martianus Capella and the Seven Liberal Arts. Columbia University Press (Records of Civilization; Wellcome Collection catalogue record).
- Findings
- Study and translation of Martianus Capella's De nuptiis Philologiae et Mercurii (5th century), the allegory in which the seven liberal arts appear as bridesmaids; the canonical seven-art list it fixed shaped medieval teaching for centuries.
- Limitations
- The original is an allegorical encyclopedia of uneven technical quality; scholarly translation choices shape how it is read.
- [9]
Hastings Rashdall (1895). The Universities of Europe in the Middle Ages, vol. 1. Clarendon Press (Internet Archive scan).
- Findings
- Classic institutional history of the medieval universities: the arts faculty as the preparatory course and theology, law and medicine as the higher faculties.
- Limitations
- Dated in parts and Eurocentric in framing; later scholarship revises many details.
- [10]
James Van Horn Melton (1988). Absolutism and the Eighteenth-Century Origins of Compulsory Schooling in Prussia and Austria. Cambridge University Press.
- Findings
- History of the compulsory-schooling movements under Frederick the Great and Maria Theresa, including Prussia's General-Landschul-Reglement of 1763; schooling served state-building, religious and economic aims together.
- Limitations
- Focused on two German-speaking states; enforcement lagged far behind the regulations.
- [15]
Martin Kemp (2006). Leonardo da Vinci: The Marvellous Works of Nature and Man. Oxford University Press (rev. ed.).
- Findings
- Shows how Leonardo's painting, anatomy, engineering and natural philosophy formed one inquiry grounded in observation and drawing; he described himself as a man without (Latin) letters.
- Limitations
- Leonardo's notebooks were largely unpublished in his lifetime; his direct influence on science was limited.
- [21]
Josef Pieper (trans. Alexander Dru) (1948 (German); 1952 (English)). Leisure: The Basis of Culture. Faber & Faber / Pantheon, trans. with an introduction by T. S. Eliot (WorldCat record).
- Findings
- Argues that the "world of total work" reduces people to functionaries; leisure, as receptive contemplation and celebration, is the root of culture and of the liberal arts.
- Limitations
- A Thomist and explicitly religious argument (some readers argue the English title overstates the thesis, which ties leisure to worship); its sociology of work is impressionistic.
- [22]
Hannah Arendt (1954/1961). The Crisis in Education (in Between Past and Future). Penguin Classics.
- Findings
- Education is where adults decide whether they love the world enough to take responsibility for it, and love children enough to prepare them to renew it, without dictating what they will make of it.
- Limitations
- A philosophical essay prompted by American school controversies of the 1950s, not an empirical study.
- [23]
Hannah Arendt (1958). The Human Condition. University of Chicago Press.
- Findings
- Distinguishes labour (sustaining life), work (making a durable world) and action (appearing and acting together in public); warns of a society of labourers about to be freed from labour, which knows no higher activities.
- Limitations
- Conceptual distinctions that are hard to operationalise; its account of the labour movement is contested.
- [24]
Ivan Illich (1971). Deschooling Society. Harper & Row (Internet Archive scan).
- Findings
- Criticises the confusion of schooling with education and of credentials with competence; proposes "learning webs" that connect learners to resources, skill exchanges and peers.
- Limitations
- A polemic; its proposals were influential as critique but rarely implemented at scale.
- [27]
Martha C. Nussbaum (2010). Not for Profit: Why Democracy Needs the Humanities. Princeton University Press.
- Findings
- Argues that education for economic growth alone neglects the critical thinking, imagination and empathy that democratic citizenship requires.
- Limitations
- Normative argument; evidence for the civic effects of humanities education is mostly indirect.
- [28]
Amartya Sen (1999). Development as Freedom. Oxford University Press.
- Findings
- The capability approach: judge development by the real freedoms people have to do and be what they have reason to value, not by income alone.
- Limitations
- Deliberately leaves the list of capabilities open, which makes measurement contested.
- [33]
Robin Wall Kimmerer (2013). Braiding Sweetgrass: Indigenous Wisdom, Scientific Knowledge and the Teachings of Plants. Milkweed Editions.
- Findings
- A botanist and Potawatomi citizen shows how Indigenous knowledge and scientific ecology can complement each other, with reciprocity and responsibility built into knowing.
- Limitations
- Essayistic; draws on particular nations' traditions that should not be generalised as "Indigenous knowledge" in the singular.
- [34]
Francisco J. Varela, Evan Thompson and Eleanor Rosch (1991/2016). The Embodied Mind: Cognitive Science and Human Experience (rev. ed.). MIT Press.
- Findings
- Founding text of enactive cognitive science: cognition arises from the history of a body acting in a world; it also brings Buddhist contemplative traditions into dialogue with cognitive science.
- Limitations
- Theoretical framework; empirical support for strong "embodied" claims varies across domains.
- [35]
Michael Polanyi (1966). The Tacit Dimension. University of Chicago Press.
- Findings
- We can know more than we can tell: skilled knowing rests on tacit, embodied integration that resists full articulation.
- Limitations
- Philosophical argument; how much tacit knowledge machine learning can capture is now an open empirical question.
- [36]
Richard Sennett (2008). The Craftsman. Yale University Press.
- Findings
- Craftsmanship, the desire to do a job well for its own sake, links hand and head; skill develops through repetition, resistance and correction.
- Limitations
- Interpretive social theory built on historical examples, not systematic measurement.
- [39]
Marie Jahoda, Paul F. Lazarsfeld and Hans Zeisel (1933 (English 1971)). Marienthal: The Sociography of an Unemployed Community. Aldine (1971); Transaction/Routledge reprint.
- Findings
- Mass unemployment in an Austrian village led not to revolt but to resignation and the collapse of time structure: evidence that employment provides more than income.
- Limitations
- Involuntary unemployment in poverty and stigma differs from a funded, chosen life without wage labour.
- [55]
Daron Acemoglu and Simon Johnson (2023). Power and Progress: Our Thousand-Year Struggle over Technology and Prosperity. PublicAffairs.
- Findings
- Historical argument that technology's gains are shared only when institutions and countervailing power direct them; the direction of technology is a choice.
- Limitations
- Broad historical synthesis; contested by economists who see stronger automatic diffusion of gains.
Peer-reviewed article
- [37]
Tyler J. VanderWeele (2017). On the Promotion of Human Flourishing. PNAS 114(31): 8148–8156.
- Findings
- Proposes five domains of flourishing (happiness and life satisfaction; health; meaning and purpose; character and virtue; close social relationships) plus financial and material stability, and reviews pathways such as family, work, education and religious community.
- Limitations
- Mostly observational evidence from high-income countries; causal claims are limited.
- [38]
Richard M. Ryan and Edward L. Deci (2000). Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being. American Psychologist 55(1): 68–78.
- Findings
- Autonomy, competence and relatedness are basic psychological needs; environments that support them foster intrinsic motivation and well-being.
- Limitations
- Cross-cultural generality of the needs has been debated, though much later research supports it.
- [40]
Robert J. Sternberg (1998). A Balance Theory of Wisdom. Review of General Psychology 2(4): 347–365.
- Findings
- Wisdom is the application of tacit knowledge, guided by values, to balance intrapersonal, interpersonal and extrapersonal interests over the short and long term toward a common good; argues wisdom can be taught.
- Limitations
- Theory-driven; evidence that school programmes increase wisdom is limited.
- [41]
Igor Grossmann, Nic M. Weststrate, Monika Ardelt et al. (2020). The Science of Wisdom in a Polarized World: Knowns and Unknowns. Psychological Inquiry 31(2): 103–133.
- Findings
- Expert consensus centres wisdom on perspectival metacognition (intellectual humility, recognising uncertainty, considering diverse perspectives, integrating them) in service of moral aims.
- Limitations
- The authors themselves list open questions about measurement, cultural variation and trainability.
- [42]
Laura Maska, Dimitrios Kalamaras and Charalambos Tsekeris (2026-07-10). The Power–Wisdom Gap: Reframing Higher Education for Human Flourishing in the Age of Artificial Intelligence. Sustainability 18(14): 7076.
- Findings
- Argues that higher education's central challenge is shifting from knowledge scarcity to power abundance; the deficit is one of formation, not of knowledge. Proposes "flourishing stewardship" as a first principle and a Flourishing Stewardship Transformation Model with six formation capacities and five institutional levers.
- Limitations
- Conceptual design with a systematised scoping review: it offers propositions and indicators for testing, not empirical evidence that the model improves outcomes. Focused on higher education.
- [46]
I. J. Good (1965). Speculations Concerning the First Ultraintelligent Machine. Advances in Computers 6: 31–88.
- Findings
- The original "intelligence explosion" argument: a machine that can design better machines could trigger recursive self-improvement.
- Limitations
- A speculative argument, not a forecast with evidence; timing claims in the paper proved wrong.
- [49]
Lisanne Bainbridge (1983). Ironies of Automation. Automatica 19(6): 775–779.
- Findings
- The more automated a system, the more crucial and the harder the human supervisor's role: operators lose the practised skill they need precisely when automation fails.
- Limitations
- Based on industrial process control; its ironies are argued, not measured, though widely confirmed since.
- [50]
Raja Parasuraman and Dietrich H. Manzey (2010). Complacency and Bias in Human Use of Automation: An Attentional Integration. Human Factors 52(3): 381–410.
- Findings
- Reviews evidence that people over-rely on automated aids (automation bias and complacency), affecting experts as well as novices, and that training alone only partly prevents it.
- Limitations
- Studies mostly predate generative AI; how the effects scale to language-model agents is still being measured.
- [51]
Duri Long and Brian Magerko (2020). What is AI Literacy? Competencies and Design Considerations. Proceedings of CHI 2020.
- Findings
- Defines AI literacy as competencies that let people critically evaluate AI, communicate and collaborate with it, and use it as a tool; proposes 17 competencies.
- Limitations
- Predates large language model agents; the competencies are a literature synthesis, not validated outcomes.
- [53]
David H. Autor (2015). Why Are There Still So Many Jobs? The History and Future of Workplace Automation. Journal of Economic Perspectives 29(3): 3–30.
- Findings
- Automation substitutes for some tasks but complements others, raising demand for labour that is hard to automate; the strongest mainstream case against simple job-loss narratives.
- Limitations
- Written before generative AI, whose task coverage is broader than previous automation.
- [54]
Daron Acemoglu and Pascual Restrepo (2019). Automation and New Tasks: How Technology Displaces and Reinstates Labor. Journal of Economic Perspectives 33(2): 3–30.
- Findings
- Labour demand depends on a race between automation that displaces workers from tasks and the creation of new tasks that reinstates them; in recent US decades displacement has outpaced reinstatement.
- Limitations
- Task-based accounting framework; whether new-task creation can keep pace with general-purpose AI is unresolved.
Primary research (institute report, not peer-reviewed)
- [43]
Megan Kinniment, Manish Shetty, Brian Goodrich, Matt Putz et al. (METR) (2026-11-20). AI R&D-relevant model capabilities are improving rapidly and are getting harder to measure. METR blog.
- Findings
- Combined TH1.1 and TH2 50% time horizons double roughly every 3.61 months (95% CI 2.08–4.94). Naive extrapolation reaches the task set's 200-hour ceiling around July 2027. Models often match or beat skilled humans on coding and optimisation tasks with strong feedback, but appear to have worse judgment than experts where feedback loops are long or solutions must generalise. The composite METR Capability Index shows no obvious general acceleration.
- Limitations
- Task horizons measure completion of benchmark tasks at a 50% success threshold, not general intelligence or job replacement. Human baselines directly support only a minority of time estimates; METR says its measures are becoming harder to interpret as tasks saturate.
- [44]
Thomas Kwa, Ben West, Joel Becker et al. (METR) (2025-03-19). Measuring AI Ability to Complete Long Software Tasks. METR blog / arXiv:2503.14499.
- Findings
- Introduces the "time horizon" metric: the length of task, measured by how long it takes skilled humans, that a model completes with 50% reliability. Found a doubling time of around seven months over 2019–2025.
- Limitations
- Software and research-engineering tasks only; tasks are cleaner than real work; reliability at higher thresholds (e.g. 80%) is much lower.
- [45]
Joel Becker, Nate Rush, Elizabeth Barnes and David Rein (METR) (2025-07-10). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. METR blog / arXiv:2507.09089.
- Findings
- In a randomised trial, experienced developers working in their own large repositories took 19% longer with early-2025 AI tools, while believing the tools had sped them up.
- Limitations
- Sixteen developers in mature codebases with early-2025 tools; METR itself warns against generalising, and later redesigned the study.
- [48]
Katja Grace, Harlan Stewart, Julia Fabienne Sandkühler et al. (2024). Thousands of AI Authors on the Future of AI. arXiv:2401.02843 (later in the Journal of Artificial Intelligence Research).
- Findings
- Survey of 2,778 published AI researchers; aggregate forecasts for full automation of all occupations sit far later than for "high-level machine intelligence", and estimates are highly sensitive to question framing.
- Limitations
- Expert opinion, not measurement; framing effects show how unstable these forecasts are.
Scholarly reference work
- [5]
Stanford Encyclopedia of Philosophy (various authors) (Rev. 2022). Medieval Philosophy. Stanford Encyclopedia of Philosophy.
- Findings
- Overview of the medieval intellectual tradition, including the transmission of the liberal arts through Boethius and Cassiodorus, the cathedral schools and the arts faculty of the universities.
- Limitations
- A philosophical overview; institutional history of schooling is treated only in passing.
- [6]
John Marenbon (Rev. 2021). Anicius Manlius Severinus Boethius. Stanford Encyclopedia of Philosophy.
- Findings
- Boethius (c. 480–524) wrote handbooks on arithmetic and music (and probably geometry), and his De institutione arithmetica uses the term quadrivium for the four mathematical disciplines.
- Limitations
- Reference entry; several attributions of Boethian mathematical works remain debated.
- [11]
Encyclopedia.com (Major Acts of Congress and related entries) (1852 (entry undated)). An Act Concerning the Attendance of Children at School (Massachusetts, 1852). Encyclopedia.com.
- Findings
- Massachusetts' 1852 act was the first US statewide law requiring school attendance; by 1918 every state had one.
- Limitations
- A reference summary; the original law was weakly enforced.
- [12]
UK Parliament (1870 (page undated)). The 1870 Education Act. UK Parliament, Living Heritage.
- Findings
- The Elementary Education Act 1870 created school boards to provide elementary schooling across England and Wales where voluntary provision was lacking.
- Limitations
- Institutional summary; attendance became compulsory only through later acts (1880).
- [29]
Jeffrey Riegel (Rev. 2013). Confucius. Stanford Encyclopedia of Philosophy.
- Findings
- Confucian learning joins textual study with ritual practice and self-cultivation; the Zhou-era "six arts" (rites, music, archery, charioteering, calligraphy and mathematics) joined bodily, aesthetic and intellectual formation.
- Limitations
- Reference entry; much of what is attributed to Confucius was compiled by later disciples.
- [30]
Therese-Anne Druart (Rev. 2020). al-Farabi. Stanford Encyclopedia of Philosophy.
- Findings
- Al-Farabi (c. 870–950) wrote an Enumeration of the Sciences that classifies the sciences of language, logic, mathematics (including music, astronomy, optics and mechanics), physics, metaphysics and political science.
- Limitations
- Reference entry; the detailed classification is summarised rather than reproduced.
- [31]
Encyclopaedia Britannica (various authors) (Rev. 2025). Ibn Khaldun. Encyclopaedia Britannica.
- Findings
- Ibn Khaldun's Muqaddimah (1377) treats crafts and sciences as products of social organisation, and analyses how knowledge, habit and teaching are transmitted in civilisations.
- Limitations
- Reference entry.
- [61]
The Post-Labor Wiki (2026). Post-Labor Wiki: a source-first guide to post-labor economics. Sister project.
- Findings
- Evidence pages on labour share, automation, cash-transfer trials, ownership models and Shapiro's framework, with empirical, forecast and normative claims kept apart.
- Limitations
- A secondary guide; follow its citations to the primary sources.
Intergovernmental report
- [25]
Edgar Faure et al. (UNESCO) (1972). Learning to Be: The World of Education Today and Tomorrow. UNESCO.
- Findings
- Made lifelong education the organising idea for educational reform and argued for the "complete" person rather than training for a single job.
- Limitations
- An aspirational policy report, not evaluated evidence.
- [26]
Jacques Delors et al. (UNESCO) (1996). Learning: The Treasure Within. UNESCO.
- Findings
- Proposes four pillars of education: learning to know, to do, to live together and to be.
- Limitations
- Aspirational and widely cited, but its pillars are broad and loosely operationalised.
Commentary and scenario writing
- [47]
Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland and Romeo Dean (2025-04-03). AI 2027. AI Futures Project.
- Findings
- A detailed scenario in which automated AI research drives rapid capability gains in 2027, with branching outcomes depending on governance choices.
- Limitations
- An explicit scenario built on judgmental forecasts; the authors present it as one plausible path, not a prediction.
- [56]
David Shapiro (2025-10-31). The "Post-Labor Enterprise". Substack.
- Findings
- Argues that AI and robotics refactor rather than abolish the firm: labour drops out as the main input, while land, capital and liability remain; scarcity shifts to physical bottlenecks of mass, distance, time and heat, and ownership of infrastructure becomes a deep moat.
- Limitations
- Scenario and strategy writing on a newsletter, not peer-reviewed measurement.
- [57]
David Shapiro (2025-07-19). Post-Labor Economics pt. I: The Rise of Automation. Substack.
- Findings
- Frames automation as proceeding wherever machines are better, faster, cheaper or safer; traces the historical shedding of agricultural, industrial and service labour; distinguishes cognitive "hyper-abundance" from post-scarcity and calls for new distribution institutions.
- Limitations
- Commentary; several figures in the post are unsourced, and physical limits are treated as engineering problems rather than measured.
- [58]
David Shapiro (2024-11-17). What do I mean when I say "Post-Labor Economics" anyways?. Substack.
- Findings
- Defines post-labor economics as the decoupling of economic growth from human labour; names an "economic agency paradox" (less wage income, less demand); rejects post-scarcity because physics and logistics still bind; proposes investment-based participation.
- Limitations
- Commentary; its tokenisation proposals are speculative and untested.
- [59]
David Shapiro (2026-01-22). These Two Things Will DESTROY CAPITALISM (The Structural End of Capitalism: Why AI Breaks the Wage-Labor Link). Substack.
- Findings
- Argues that when wages stop distributing purchasing power, a "realisation crisis" follows and the default drifts toward techno-feudalism; proposes non-wage demand through dividends on machine rents and "sovereign equity" in automated infrastructure.
- Limitations
- Commentary and scenario; the URL slug differs from the published title.
- [60]
David Shapiro (2026-01-29). Why AI is slowing down in 2026. Substack.
- Findings
- Argues that 2026 frictions are physical and operational (grid interconnection, transformers, memory, packaging, deployment) rather than regulatory; calls the period a "digestion phase" before acceleration resumes around 2028.
- Limitations
- Commentary aggregating third-party figures that we have not independently verified.