Coherence Without Certainty
HUM×AI=n
The problem with having worked across so many fields is that it becomes difficult to describe the work at all. Search engines want a category, institutions want a discipline, and large datasets want a label that allows one person, one project, or one idea to be placed somewhere among billions of others. My work does not always fit neatly into those categories. Network mathematics sits beside education, artificial intelligence beside governance, open source data beside questions of identity, work, responsibility, and extraction. From the outside, this can look like a collection of mismatched ideas. From where I stand, the connections have become increasingly difficult to ignore.
Every system runs on assumptions that no one inside it can see. That is what makes them assumptions. A price signal, a curriculum, a regulation, a recommendation feed: each encodes a decision about what matters that was made once, by someone, under conditions that may no longer hold, and then forgotten. The forgetting is the mechanism. Nobody has to defend an assumption that has become the shape of the room.
This essay asks what follows when we take that observation seriously. I approach it through the different instruments I have used to understand how systems hold together and how they begin to come apart: network mathematics and the mathematics of decision, open source data tools, classrooms and the changing meaning of work, regional strategy and governance, and philosophical questions of responsibility, identity, and extraction. Taken separately, these subjects can look mismatched, and much of my own work began that way, as individual questions, projects, arguments, and observations arising in very different contexts. Taken together, however, I increasingly see them as different ways of examining the same underlying problem: how do we develop systems that can learn, adapt, and correct themselves as the assumptions beneath them evolve?
The connection I am trying to develop is therefore less about creating a single theory than about bringing together a set of values that have emerged across this work: human agency, distributed possibility, accountability, openness to correction, and the conviction that technological capability should expand what people are able to become rather than simply make existing systems more powerful. I want to understand where these values reinforce one another, where they come into tension, and whether they can provide a useful foundation for the way I develop and think about artificial intelligence.
AI makes this effort more consequential because the systems we are building are increasingly participating in the very domains through which we organize human life. They can influence how we learn, decide, work, create, govern, and understand ourselves, which means that their development inevitably carries assumptions about what matters, whose interests count, what should be optimized, and what kinds of futures should remain possible.
What follows is therefore an attempt to bring these strands of my work into the same frame and examine what they reveal when considered together. Some of the ideas are established through evidence, some remain hypotheses, and some are tensions within my own thinking that I have not resolved. I want to keep those distinctions visible because the purpose of bringing the work together is not to make it appear more complete than it is, but to see whether these ideas can become more useful when placed in conversation with one another, particularly as I continue trying to develop AI that expands human possibility and benefits people more broadly.
I. A civilization is several systems moving at different speeds
Think of a civilization as a set of interconnected layers, each moving at its own pace. Tools can change within months, while business models and incentives often take years to evolve. Laws and institutions may take decades to shift, while schools and culture are shaped over generations. Beneath all of these layers are the deeper assumptions that determine what we consider success, who we recognize as a participant, what we believe is worth protecting, and what we understand a good life to be. These assumptions change more slowly than anything built on top of them, yet they are often the least visible. Precisely because they are so deeply embedded in the systems around us, we rarely recognize them as assumptions at all.
The layers are interdependent. The slow layers set the objectives that guide the fast ones. A company optimizes what its incentives reward; those incentives reflect what an institution has decided to measure; and the institution measures what its era assumes matters. When the fast layer accelerates ahead of the slow layers, it continues optimizing the objective those layers established, even as the surrounding conditions change.
I wrote a version of this argument in Spanish this year: the persistence of divisions between countries, ideologies, and communities reveals a gap between the historical structure of power and newly emerging capacities for interconnection. [9] Although the argument may initially seem tied to an earlier moment, I believe it becomes relevant when viewed alongside the historical work of economist Carlota Perez. Her framework offers a useful account of this dynamic. In her analysis, long waves of technological change reflect a tension in the internal coherence of a system: the techno economic sphere can transform much faster than the socio institutional sphere, creating a mismatch that has historically lasted two or three decades. Institutions tend to carry forward established practices shaped by past successes and vested interests, even as the technological environment around them changes. Perez also argues that these periods unfold through trial and error, as societies navigate profound change without proven recipes and absorb significant social costs along the way. [19] Her chapter examines an earlier information technology transition, so applying this pattern to AI remains an extrapolation. Whether the same dynamic will hold, and whether AI will intensify or alter it, remains an open question.
“Exponential” names a claim about rates. I use the word narrowly: exponential change is compounding, with growth that depends on the current level. My network model captures the idea in its simplest form: small asymmetries in influence accumulate into large concentrations of power in a few nodes. [1] The systems thinker Donella Meadows, whom I draw on, stresses that delays between action and consequence make complex systems difficult to manage. [4] Together, these ideas describe a system in which compounding growth moves faster than the loops designed to correct it, allowing second- and third-order effects to arrive before the first-order effect is understood.
What is new about AI? Earlier technologies mostly operated within human assumptions, making us faster at what we already meant to do. AI systems increasingly participate in forming those assumptions: they shape what we see, whom we follow, how we draft an argument, and what counts as a good answer. I have argued that AI amplifies whatever it is given, so an unexamined assumption can become faster and more persuasive while retaining the same degree of correctness. [6][7] My claim here is that the fast layer now reaches directly into the slowest one.
However, I must note that the evidence remains partial. A 2026 Nature experiment on X, a social media platform that has become increasingly difficult to distinguish from the information environments it helps shape, found that an algorithmic feed shifted political opinions, while switching the algorithm off left those shifts intact, apparently because the algorithm had changed whom people chose to follow. [25] A survey of 319 knowledge workers found that greater confidence in generative AI was associated with less critical thinking, while greater self-confidence was associated with more, though the data were self-reported. [26] An earlier large study with Meta found no significant effect of a chronological feed on polarization over three months, while critics noted that the platform changed its algorithm during the study. [24] These studies measure downstream effects rather than assumptions directly. The claim that AI reaches the assumption layer therefore remains a hypothesis with suggestive support rather than an established result.
II. Three tests for coherence
Coherence is often imagined as agreement: economies, technologies, and institutions pointing in the same direction. In practice, however, the pursuit of coherence has often served kings and queens, supreme leaders and political authorities, as well as celebrities and cultural figures, by creating systems in which entire populations were expected to accept a particular version of reality as unquestionable. The more completely those systems aligned around a single narrative, the harder it became for people to distinguish coherence from conformity. That distinction matters because agreement can be manufactured through authority, repetition, exclusion, or the suppression of dissent. A system can therefore appear remarkably coherent while becoming less capable of questioning itself. I would define coherence differently, through three tests that a set of systems must pass together.
Contact. Do the layers feel each other fast enough? A change in one should be visible to the others before it hardens into damage. This is a claim about feedback and rates.
Answerability. Are the goals each layer optimizes answerable to the people affected? A system can be internally consistent and still answer to no one.
Revisability. Can the assumptions under the system, and under the people in it, be surfaced and changed without a crisis? This is the test I care about most, because it is the one that lets a system discover it was wrong.
A society could pass the first two and fail the third: efficient, accountable, and yet unable to recognize what it has stopped questioning. The coherence worth seeking is therefore not agreement among systems, but an understanding of how they interact, where their assumptions come from, and whether they remain capable of responding to one another. We may have discovered a technology capable of transforming them all, but transformation alone does not tell us whether the direction is right, whether the assumptions remain valid, or whether the people affected can still challenge them.
After all, we are a species that has repeatedly enslaved other beings, treated the natural world as property, and built systems of extraction that concentrate benefits while distributing their costs across people, communities, and ecosystems. The capacity to transform a system is therefore not evidence that we understand what we are transforming, or that the transformation will serve a future worth creating. What matters is their shared capacity to be corrected.
The next three sections use the different fields of my work to examine each of these ideas and invite your examination of some of our most deeply held assumptions.
III. Contact: making one layer visible to another
Contact fails when consequences are invisible to those who could act on them. Three of my projects are, in different ways, attempts to build contact. The following are examples of what contact can look like in practice.
A heat map. HeatWatch, an open source idea of mine, uses satellite land surface temperature and vegetation data to map urban heat, score neighborhood level vulnerability, and export files that planners can use. Its stated motive is that most cities lack high resolution heat maps, neighborhood vulnerability data, and tools to evaluate cooling interventions. [15] It is a small example of a larger idea: a physical layer of heat, trees, and concrete made legible to a social layer of who lives where and who is at risk, and then to a political layer of what to fund. I should be candid about its status. The repository is still early, with not that many commits, no stars, three forks, and a README that still contains template placeholders awaiting collaboration. So I leave it to you to decide what this example demonstrates. For me, although it shows the shape of contact, the absence of proof offers you a glimpse of what becomes possible when something that was previously invisible is made visible to those who can act on it.
A governance essay. In examining regional economic systems, I found a recurring problem: policy often focuses on visible instruments such as tax incentives while leaving the deeper architecture of decision making largely untouched. Who has access to information? How are rules designed? Who can question them? Who can see where resources are going and what results they produce? These questions led me back to the work of systems thinker Donella Meadows, who argued that changing the structure through which information flows can be one of the most powerful ways to influence a complex system. [4] The remedy I proposed was therefore not simply another policy instrument, but stronger contact between decisions and the people affected by them through open contracting, public expenditure dashboards, independent audits, and more transparent institutional rules. When information moves only upward, outward, or through closed channels, consequences can remain invisible to the people with the greatest need or ability to respond. When information becomes visible, traceable, and contestable, a system gains another possibility: not merely to operate, but to learn from what its decisions produce.
Regional standards. In a federal filing, I argued that Latin America and the Caribbean could become a leading AI region through shared data standards, joint research, and capacity building. [17] A separate 2026 proposal hosted by the Stimson Center describes the region's AI landscape as fragmented and proposes pooling compute, talent, and funding so the region builds capability rather than dependency. It is a proposal, not a result, but it points to the same underlying challenge: when countries and institutions develop technological capacity in isolation, they may have less ability to shape the systems they collectively depend on. Shared standards can create contact across borders, allowing data, knowledge, infrastructure, and institutions to interact rather than develop as disconnected parts. [28]
Contact also explains a finding I keep returning to. When a technology's effects are invisible to its users, the effects compound. A recent study of one technology firm found that employees using generative AI worked faster, took on broader tasks, and extended work into more hours, often unasked. The authors call it in-progress research at a firm of about 200 people, so it cannot be generalized. [27] But it fits the mechanism: the time a tool saves is absorbed by expectations before anyone sees it happen. Making that visible is a contact problem.
IV. Answerability: who can be asked
Answerability fails in two ways: nobody is responsible, or those who benefit are not those who bear the cost.
Nobody is responsible. I argue that responsibility in a sociotechnical system depends not only on an agent's own actions but on access to information and on the agent's relations with other agents. Assuming full responsibility would mean accounting for all the effects that emerge from collective interaction. [9] Philosophers call the underlying difficulty the problem of many hands: when many people contribute to a decision, it is hard even in principle to say who is morally responsible for the result. [23] Dennis Thompson, who named it in 1980, proposed criteria of personal responsibility: a person is responsible for an outcome insofar as their actions or omissions are a cause of it and they did not act in ignorance or under compulsion. He suggested the problem can be mitigated by designing processes of responsibility. [23] I want to make a connection he did not, and which is a synthesis: if responsibility depends on what a person could know, then transparency is a precondition of answerability, not a courtesy. That is why open contracting in governments and legible heat maps belong in the same argument. You cannot answer for what you could not see.
Benefit and cost fall on different people. Acemoglu and Johnson argue that technological progress is not automatic and depends on choices, which can serve a narrow elite or become the foundation for widespread prosperity. [20] Their critics dispute some of their historical examples. Noah Smith, for instance, calls their treatment of synthetic fertilizer badly overstated. [20] The structural claim survives that dispute.
I made an adjacent argument about colonization: that the colonizer also wounds itself, corrupting its own institutions and social trust, and that AI trained on data steeped in colonial-era bias risks building a new empire of data and algorithmic power. [10] I want to hold that claim to a higher standard here. The economic literature on colonial institutions suggests a less comfortable point. Acemoglu, Johnson, and Robinson argue that extractive institutions persisted where a small ruling elite gained a lot from extraction, and where changing institutions was costly. [21] Their identification strategy has itself been challenged; Albouy reexamined the settler-mortality data. [21] Still, the implication for my argument is that extraction can be stable and even profitable for the narrow group that benefits, while the costs are diffuse or borne by others. That does not refute my point about long-run harm to the colonizer's society, which is moral and institutional as well as economic. But it means "extraction hurts the extractor" cannot be assumed to persuade those who profit from it. It is another reason coherence must be measured by whom it is coherent for.
My book makes a related point about design: platforms default to English and to urban centers, so rural, migrant, and Indigenous users are absent from the making of the tools that govern them. [18] A system built without the people it governs is unanswerable to them by construction.
V. Revisability: the capacity to change our minds, and ourselves
Revisability is where my fields of work meet most directly, because it concerns assumptions held by machines, institutions, and persons.
Identity. My theory treats identity fragmentation as harm and identity coherence as its opposite. [1] Yet an essay I wrote in 2025 says the next mindset to be dismantled is identity itself, and that those who cling to static identities will become outdated by design, while those who adapt will practice a fluid "post-ego" navigation. [11] These pull in different directions. I think they can be reconciled with a distinction: coherence is not fixity. A coherent self is one that can revise its story without losing its thread, which is revisability applied to a person. But I should also note a risk the reconciliation does not remove. For communities whose identities were suppressed by colonization, as I have written about elsewhere, [10] identity is also a resource for resistance. Asking everyone to loosen identity equally asks the least powerful to give up the most. I have not resolved this.
Decisions. In a short 2026 reflection, I proposed that decisions can be represented as nodes in a network of possibilities, with probabilities, utilities, and outcomes, and that we must learn to decide before AI begins deciding for us. [12] A booklet of mine on the same theme explores how to preserve human authority and accountability in an era of AI systems capable of producing fast, fluent, and scalable recommendations. [14] There is a genuine strength in representing decisions mathematically: it forces us to make our assumptions visible. We have to ask what outcomes we consider possible, how likely we believe they are, what we value, and what we are willing to trade against something else. But the mathematics also carries assumptions of its own, particularly the idea that the values at stake can be compared and expressed through utilities. Not everything that matters to a person, a community, or a society can necessarily be placed on the same scale. The survey finding above complicates this further: people who place greater confidence in generative AI have been found to engage in less critical thinking. [26] Learning to decide, therefore, is not simply learning how to calculate better. It also means learning when calculation clarifies a decision, when it conceals an assumption, and when a decision should not be reduced to a number at all.
Work. I have argued that a job is a structure created by an organization, while work is something larger: the human activity of creating, solving, caring, building, learning, and contributing. AI is beginning to make the distinction between these two ideas more visible. As some tasks become automated or reorganized, the possibility emerges to design work around people's capabilities and potential rather than designing people around the boundaries of existing jobs. [8] But that possibility raises a deeper question of revisability. Can the institutions built around jobs, including status, credentials, career paths, compensation, and even the way we introduce ourselves, change as quickly as the technologies that are beginning to unbundle them? This is Perez's mismatch again, but applied not only to institutions and technology, but to the meaning of a career and the role work plays in our lives. The transition also carries a danger. A recent study found that AI did not simply reduce the amount of work people performed; in the organization studied, workers became faster, took on broader tasks, and extended their working hours. [27] The freedom created by AI, therefore, does not arrive automatically. If institutions do not change with the technology, the time AI appears to give back may simply become capacity for more work.
Learning. I must be cautious and fair when applying this argument to my own work. One of my ventures builds an adaptive K 12 learning platform that uses machine learning to personalize instruction in real time. [16] My research agenda also includes rethinking personalized learning as collective liberation and examining how power structures can undermine student identity through educational technology. [3] This creates a tension within my own work. My theory identifies preference amplification as a mechanism of fragmentation, arguing that recommendation systems can narrow the range of identity resources available to users. [1] From that framework, I can infer a potential risk for adaptive education: a system that continually personalizes instruction could reinforce existing preferences or assumptions about a student rather than expose the student to a broader range of possibilities. [1][16] This concern remains an inference from my framework, rather than a conclusion established by the evidence from my platform. The distinction matters. I want to examine the possibility without turning a theoretical concern into a claim about a system for which the evidence does not yet support that conclusion.
The data our research has gathered, together with the evidence available so far, suggest that this issue deserves attention, but they do not establish that the platform produces the broader effect described by the theory; therefore, we continue to investigate it. Even on my own platform, that distinction matters. Because personalization depends on assumptions. Personalization necessarily involves assumptions about what a learner needs, how the system interprets their responses, and what it chooses to present next. The standard I want to apply is whether an adaptive system can make the assumptions behind its personalization visible, open them to examination and challenge, and allow both the learner and the teacher to change them. That is the revisability test I would apply to any adaptive learning system, including my own.
VI. How the pursuit of coherence goes wrong
Disclaimer, I do not think these risks can be designed away.
Coherent for what? Perez notes that a paradigm defines a range of possibilities, not goals. Under the previous mass-production paradigm she lists four major modes of growth: Keynesian democracy, fascism, socialism, and state developmentalism. They are profoundly different, yet they share features that stem from the same underlying logic. [19] Fit between a technology and its institutions does not tell us which society results. My own working paper asks "collective empowerment toward what ends?" and concedes that the answer needs philosophical work. [1] So far my answer is distributed power and collective responsibility. That is a value I hold, not a finding.
Coherence can lock in. Perez also describes how systems create advantages for innovations that fit them and exclusion for those that do not. [19] A system tuned to itself can become blind to radical alternatives. Regional standards could entrench a design that a later generation would need to abandon.
Coherence can centralize. The Ostrom tradition argues that multiple, overlapping, autonomous decision centers can improve efficiency, accountability, and problem-solving. Elinor Ostrom credited polycentric systems with mutual monitoring, learning, and adaptation. [22] Polycentricity has context-dependent drawbacks, so this is not a trump card. But an actor built to align everything becomes a single point of failure. The coherence worth wanting looks more like shared rules across many centers than alignment under one plan.
Audits can become orthodoxy. Revisability requires examining assumptions. Turned into a compliance exercise, assumption audits become loyalty tests, and the people best placed to question the system learn to stay quiet. The remedy I can think of is that the audit function must itself be auditable, with a protected channel for dissent.
Measurement can distort. In my model, fragmentation is a score for each node and coherence is one minus that score. Nothing in it represents diversity of views across people. [1] So the model cannot tell a plural society of coherent selves from a uniform one, and a coherence metric optimized directly could reward uniformity. That is my reading of the model, and I am not certain it matches how I would defend it.
Capture. New institutions built to create coherence become prizes for the interests they were meant to check. [4]
VII. Where I may be wrong
The most useful sentence in a body of work is often the one that says where it might fail. As far as I can discern, mine are these.
The four forces may not carve reality. Greed, fear, influence, and responsibility are moral-psychological categories that I have used as structural variables. Whether they hold up empirically is exactly what my working paper says must still be shown, through scale development, discriminant validity against related constructs, and replication. [1]
The simulation confirms its own assumptions. It uses ten nodes with illustrative parameters. [1] The paper is candid that the parameters are uncalibrated, the graph is static, the dynamics are deterministic, and the intervention is an external input. [2] A model that assumes interventions reduce fragmentation will find that they do.
My reconciliation of accelerant and amplifier leans on the least verified claims. For a while my papers said AI accelerates fragmentation and my posts said AI could expose our assumptions. A 2026 paper reconciles them with a time-indexed argument: accelerant now, possible resolver later, with the danger concentrated in a transition. [2] That paper says its convergence hypothesis is structurally plausible but empirically undemonstrated, that its transition mechanism is underspecified, and that it faces a circularity. [2] Some of its propositions are philosophically contested, and one, recasting natural disruptions as recalibrations, could be misread as rationalizing harm. [2] I think the coherence argument here is stronger when it does not depend on them.
Mindset can be a way of not listening. A recent post traces fears about AI to survival-shaped cognition that notices threats before opportunities. [6] An earlier Spanish essay locates a social breakdown partly in individual impulsivity and offers self-control as the lever, though it also notes that digital environments encourage impulsivity. [13] My papers, by contrast, place the cause in structures of power and platform incentives. [1] These are different theories of change. If the bottleneck is mindset, education is the lever. If it is structure, institutions are. And a mindset explanation can be turned against people whose fears are well founded. Explaining a worry is not answering it.
Empowerment and competitiveness may conflict. My theory draws on liberation and postcolonial traditions, [1][3] while my regional filing frames part of the case through industrial competitiveness and alignment with the interests of the United States and its closest allies. [17] Note: I have not shown these are reconcilable.
Thought leadership is not evidence. A published reviewer said, as I reported, that my book leans heavily on conceptual vision, and readers wanted more industry-specific case studies. [18] The remedy is not a better narrative. It is measurement, which is why projects like a heat map matter more than they look.
VIII. Three illustrations, not predictions
To make the tests concrete, here are three thought experiments. I do not know that any of them works.
A neighborhood. A city publishes a heat map that overlays temperature, tree cover, and residents' vulnerability. Contact:planners and residents see the same picture. Answerability: residents help decide which interventions to fund and whose data counts. Revisability: the map is re-run each year and the metrics are retired if they fail to predict harm. The risk: a vulnerability score can be used to target help or to justify neglect. Which happens depends on governance, not on the map. [4][15]
A school. An adaptive tool personalizes instruction, and assessment rewards the reasoning behind an answer. Contact:teachers see how the tool models each student. Answerability: students and parents can contest the tool's assumptions about them. Revisability: the school reviews annually which assumptions its tools and assessments embed. The risk: the time AI returns to teachers is absorbed by more administration, unless protected time is written into institutional rules, and the tool narrows a student's range unless it is designed to widen it. [3][5][27]
A regional corridor. Governments, universities, and firms share data and interoperability standards for a common service, such as logistics. Contact: incident data is pooled and visible. Answerability: communities affected by the infrastructure have a voice in its rules, not only as customers. Revisability: standards are kept minimal and scheduled for sunset review, so the arrangement can be dismantled by design and not by failure. The risk: an early standard becomes a lock-in that later builders resent, or the pooled authority becomes the prize for capture. [4][19][28]
IX. The assumption that will need dismantling
What if the frameworks I propose today become the assumptions a future generation must dismantle?
This essay's own assumptions are easy to list. That systems can be described as layers with speeds. That mismatch, not malice, explains much of what goes wrong. That contact, answerability, and revisability are the right tests. That correction is more valuable than agreement. That flourishing can be recognized from where we stand. Each may be partial, located in a particular era and a particular vantage point, or wrong. The last is the one I can least defend, and it is the one on which the others rest.
So the essay applies its own third test to itself. If revisability is what makes a system worth trusting, then a body of thought about coherence should be built to be revised: its models calibrated against data, its categories challenged by people who do not share its vocabulary, its claims retired when they fail. My earlier work says that whether the theory becomes science depends on measurement, calibration, peer review, and replication. [1][2] I take that sentence to be the most important one in it.
What I can defend is modest. Economic possibility, technological capability, and social progress will not reinforce one another by default. Whether they do is a matter of design, power, and patience, done by people who expect to be corrected. And coherence is compatible with very different kinds of societies, so the question of what we are coherent for remains open and remains political. If this work has a legacy, it may be a set of conditions that lets later generations see farther than we can, including seeing what we got wrong.
Transformation is not about proving that we were right. It is about creating a future capable of becoming better than what we currently know how to imagine.
Sources :
Working paper, "The Dismantled Mindset in the Age of Artificial Intelligence" (Zenodo, 30 May 2026). Read in full. https://zenodo.org/records/20466800
Working paper on superintelligence and convergence (2026). Read in full. https://www.fcquiles.com/superintelligence-dismantled-mindset-convergence
SSRN author page with current research directions. Read. https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7228498
Blog essay on Puerto Rico and Industry 4.0 (17 Sept 2026). Read in full. https://www.fcquiles.com/post/from-dependency-1-0-to-innovation-4-0-rewiring-a-regional-economic-system
Blog, "What If AI's Greatest Gift to Humanity Is Time to Think?" (7 Sept 2026). Read. https://www.fcquiles.com/post/what-if-ai-s-greatest-gift-to-humanity-is-time-to-think
Blog, "AI Will Amplify Humanity" (28 Aug 2026). Read. https://www.fcquiles.com/post/ai-will-amplify-humanity-the-question-is-what-part-of-humanity-we-give-it
Blog, "Was Kahneman Preparing Us for AI?" (6 Sept 2026). Read. https://www.fcquiles.com/post/was-kahneman-preparing-us-for-ai
Blog, "The Future of Work Is a Question of Human Possibility" (17 Aug 2026). Read. https://www.fcquiles.com/post/the-future-of-work-is-a-question-of-human-possibility
Spanish essay on AI redefining the limits of the past (3 Apr 2026). Read. https://www.fcquiles.com/post/la-inteligencia-artificial-es-humanidad-redefiniendo-los-límites-del-pasado
Blog, "Colonization: How to Shoot Yourself in the Foot" (29 Sept 2025). Read. https://www.fcquiles.com/post/colonization-how-to-shoot-yourself-in-the-foot
Blog, "The Day the Algorithm Disrupted Our Mindsets" (26 Sept 2025). Read. https://www.fcquiles.com/post/the-day-the-algorithm-disrupted-our-mindsets
Blog reflection on the mathematics of decisions (24 Feb 2026). Read. https://www.fcquiles.com/post/be-kind-be-still-but-learn-to-decide-how-decisions-can-be-explained-mathematically-and-define-t
Spanish essay on self-control and social breakdown (5 Oct 2025). Read. https://www.fcquiles.com/post/por-qué-estamos-rotos-como-sociedad-el-autocontrol-como-clave-de-la-convivencia
Product page for the booklet on learning to decide. Description only. https://www.fcquiles.com/product-page/learning-to-decide-in-the-age-of-ai-mastering-choices-before-machines-do
HeatWatch repository README. Read. https://github.com/castroquiles/HeatWatch
GitHub profile README (project descriptions). Read. https://github.com/castroquiles
Federal RFI response on Latin America and the Caribbean (2025). Read in full. https://files.nitrd.gov/90-fr-9088/Felipe-Castro-Quiles-GENIA-AI-RFI-2025.pdf
The book: Dismantled: A Theory of Broken Mindsets—A Blueprint of Infinite Futures by FC Quiles https://www.amazon.com/dp/B0FXFS8TVD
External
Perez, C., "Technological Revolutions, Paradigm Shifts and Socio-Institutional Change" (2004; from a 1998 original). Read in full. https://carlotaperez.org/wp-content/downloads/publications/organizational-change/TRs_TEP_shifts_and_SIF_ch.pdf
Acemoglu and Johnson, Power and Progress (2023). Publisher and review summaries only. https://www.hachettebookgroup.com/titles/daron-acemoglu/power-and-progress/9781541702547/ ; https://www.noahpinion.blog/p/book-review-power-and-progress
Acemoglu, Johnson, and Robinson (2001), "The Colonial Origins of Comparative Development," American Economic Review; Albouy reexamination (2004). Abstracts and summaries only. https://www.aeaweb.org/articles?id=10.1257%2Faer.91.5.1369 ; https://eml.berkeley.edu/%7Ewebfac/dromer/e237_F04/albouy.pdf
Ostrom tradition on polycentric governance. Secondary summaries only. https://www.thecgo.org/books/the-environmental-optimism-of-elinor-ostrom/chapter-2-self-governance-polycentricity-and-environmental-policy/ ; https://en.wikipedia.org/wiki/Bloomington_school
Thompson, D. F. (1980), "The Moral Responsibility of Public Officials: The Problem of Many Hands," American Political Science Review 74(4), 905-916. Abstract and summaries only. https://www.cambridge.org/core/journals/american-political-science-review/article/abs/moral-responsibility-of-public-officials-the-problem-of-many-hands/39DD3FAB7BF7DC7A242407143674F22B
Guess et al. (2023), Science, and critique coverage. Abstract and news coverage read. https://pubmed.ncbi.nlm.nih.gov/37498999/ ; https://www.science.org/content/article/study-found-facebook-algorithm-didnt-promote-political-polarization-critics-doubt
Gauthier et al. (2026), "The political effects of X's feed algorithm," Nature. Abstract only. https://www.nature.com/articles/s41586-026-10098-2
Lee et al. (2025), CHI 2025. Abstract and summaries only. https://dl.acm.org/doi/10.1145/3706598.3713778
Ranganathan and Ye (2026), Harvard Business Review. Summary page read; body paywalled. https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it
Stimson Center regional AI cooperation blueprint (2026). Search summary only; the site blocked full access. https://www.stimson.org/2026/regional-cooperation-in-artificial-intelligence-a-blueprint-for-international-cooperation-in-latin-america-and-the-caribbean/
Comments