Freeze-Dried Text Experiment

It is like instant coffee, or a shrunken pear: too dry to eat, but OK if you add water.  Meet "freeze-dried text" – concentrated idea nuggets waiting to be expanded by AI. Copy everything below this paragraph into any AI and watch as each transforms into real text. Caution: AI will hallucinate some references. Remember to type "NEXT" after each expansion to continue. Avoid activating any deep search features – it will slow everything down. This could be how we communicate soon – just the essence of our thoughts, letting machines do the explaining. Perhaps the textbooks of the future will be written that way. Note, the reader can choose how much explanation they really need - some need none, others plenty. So it is a way of customizing what you read. 

Mother Prompt

Expand each numbered nugget into a detailed academic paper section (approximately 500 words) on form-substance discrimination (FSD) in writing education. Each nugget contains a concentrated meaning that needs to be turned into a coherent text.

Maintain a scholarly tone while including:

• Theoretical foundations and research support for the claims. When citing specific works, produce non-hallucinated real reference list after each nugget expansion.  

• Practical implications with concrete examples only where appropriate.

• Nuanced considerations of the concept's complexity, including possible objections and need for empirical research. 

• Clear connections to both cognitive science and educational practice.

• Smooth transitions that maintain coherence with preceding and following sections

Expand nuggets one by one, treating each as a standalone section while ensuring logical flow between sections. Balance theoretical depth with practical relevance for educators, students, and institutions navigating writing instruction in an AI-augmented landscape. Wait for the user to encourage each next nugget expansion. Start each Nugget expansion with an appropriate Subtitle 

Nuggets

1. Form-substance discrimination represents a capacity to separate rhetorical presentation (sentence structure, vocabulary, organization) from intellectual content (quality of ideas, logical consistency, evidential foundation), a skill whose importance has magnified exponentially as AI generates increasingly fluent text that may mask shallow or nonsensical content.
2. The traditional correlation between writing quality and cognitive effort has been fundamentally severed by AI, creating "fluent emptiness" where writing sounds authoritative while masking shallow content, transforming what was once a specialized academic skill into an essential literacy requirement for all readers.
3. Cognitive science reveals humans possess an inherent "processing fluency bias" that equates textual smoothness with validity and value, as evidenced by studies showing identical essays in legible handwriting receive more favorable evaluations than messy counterparts, creating a vulnerability that AI text generation specifically exploits.
4. Effective FSD requires inhibitory control—the cognitive ability to suppress automatic positive responses to fluent text—paralleling the Stroop task where identifying ink color requires inhibiting automatic reading, creating essential evaluative space between perception and judgment of written content.
5. The developmental trajectory of FSD progresses from "surface credibility bias" (equating quality with mechanical correctness) through structured analytical strategies (conceptual mapping, propositional paraphrasing) toward "cognitive automaticity" where readers intuitively sense intellectual substance without conscious methodological application.
6. Critical thinking and FSD intersect in analytical practices that prioritize logos (logical reasoning) over ethos (perceived authority) and pathos (emotional appeal), particularly crucial for evaluating machine-generated content that mimics authoritative tone without possessing genuine expertise.
7. The "bullshit detection" framework, based on Frankfurt's philosophical distinction between lying (deliberately stating falsehoods) and "bullshitting" (speaking without concern for truth), provides empirical connections to FSD, revealing analytical reasoning and skeptical disposition predict resistance to pseudo-profound content.
8. Institutional implementation of FSD requires comprehensive curricular transformation as traditional assignments face potential "extinction" in a landscape where students can generate conventional forms with minimal intellectual engagement, necessitating authentic assessment mirroring real-world intellectual work.
9. Effective FSD pedagogy requires "perceptual retraining" through comparative analysis of "disguised pairs"—conceptually identical texts with divergent form-substance relationships—developing students' sensitivity to distinction between rhetorical sophistication and intellectual depth.
10. The pedagogical strategy of "sloppy jotting" liberates students from formal constraints during ideation, embracing messy thinking and error-filled brainstorming that frees cognitive resources for substantive exploration while creating psychological distance facilitating objective evaluation.
11. Students can be trained to recognize "algorithmic fingerprints" in AI-generated text, including lexical preferences (delve, tapestry, symphony, intricate, nuanced), excessive hedging expressions, unnaturally balanced perspectives, and absence of idiosyncratic viewpoints, developing "algorithmic skepticism" as distinct critical literacy.
12. The "rich prompt technique" for AI integration positions technology as writing assistant while ensuring intellectual substance comes from students, who learn to gauge necessary knowledge density by witnessing how vague AI instructions produce sophisticated-sounding but substantively empty content.
13. Assessment frameworks require fundamental recalibration to explicitly privilege intellectual substance over formal perfection, with rubrics de-emphasizing formerly foundational skills rendered less relevant by AI while ensuring linguistic diversity is respected rather than penalized.
14. FSD serves as "epistemic self-defense"—equipping individuals to maintain intellectual sovereignty amid synthetic persuasion, detecting content optimized for impression rather than insight, safeguarding the fundamental value of authentic thought in knowledge construction and communication.
15. The contemporary significance of FSD extends beyond academic contexts to civic participation, as citizens navigate information ecosystems where influence increasingly derives from control over content generation rather than commitment to truth, making this literacy essential for democratic functioning.

Two Reactions to AI

A batch of student essays. About a third are clearly AI-generated. Two professors—same discipline, same university, same evidence—react in diametrically opposite ways. Rodrigo sighs with relief. Jane spirals into panic.

For Rodrigo, it is almost liberating. If his students can now write coherent college-level essays with the help of machines, then he is free to teach them something more ambitious. Argument structure, epistemology, stylistic nuance—areas where automation falters. He is not naïve; he knows AI is here to stay. But rather than fight it, he welcomes the detour. Less time marking the same intro-to-critical-writing dreck, more time pushing the intellectual envelope. Lucky him.

Jane, however, reads the situation as academic apocalypse. Her course was the product of years of iteration, finely tuned to teach writing through careful scoping, scaffolding, and feedback. Skip the process, she believes, and you skip the learning. The AI is not a tool in her eyes; it is a cheat code, one that threatens to render her teaching obsolete. She starts researching detection tools, imagining a future of surveillance, suspicion, and pedagogical collapse.

These are not just personality quirks or different thresholds for academic dishonesty. What really separates them is how they understand curriculum. For Rodrigo, curriculum is plastic—something owned, revised, improved. He feels empowered to tinker. If a foundational skill can be outsourced, then the baseline has shifted, and he can raise the stakes. A change in student capability is an opportunity, not a crisis.

Jane sees curriculum differently. For her, it is an infrastructure. Complex, interdependent, and not easily re-routed. Learning outcomes, general education requirements, accreditation standards—these are not suggestions, they are fixtures. If a key skill like essay-writing becomes an unreliable indicator of mastery, the whole sequence threatens to unravel. You cannot simply skip a floor in the building and hope the roof holds.

There is a quiet tragedy here. Not because Jane is wrong—her concerns are deeply valid—but because she feels disempowered by a system she herself has worked so hard to uphold. The larger structures of academia—its bureaucracies, its resistance to rapid change—amplify the sense of threat. It is not just that students are using ChatGPT; it is that there is no institutional plan, no curricular pivot, no workflow update to guide faculty through this transition. So each professor is left to improvise, bringing their own philosophies and tolerances to bear.

And that is where the real tension lies. Technology does not just disrupt skills—it exposes fault lines in our educational ideologies. Are we guardians of a process or facilitators of progress? Should we protect the sequence, or adjust the map when shortcuts appear?

Rodrigo shrugs and walks forward. Jane looks for the brakes. But maybe it is not about who is right. Maybe the more urgent task is to build a system where professors do not have to choose between clinging to the past and embracing a future they did not ask for. Because either way, the syllabus is no longer entirely ours to write.

UPD: Thanks to my colleague Hogan Hays for his thoughtful critique of this blog. 

RAG and the Tyranny of Text

Writing and reading are, at their core, terribly inefficient. To communicate knowledge, we take  complex non-linear understanding and flatten it into a linear string of symbols—words, sentences, paragraphs—then expect someone else to decode those symbols one by one to reconstruct the original meaning. For every piece of information useful to us in a particular moment, we probably read thousands of unnecessary words. Laws, academic research, instruction manuals—entire professions exist solely to interpret and summarize the large texts, and find the bits useful for a particular case.

We are so accustomed to this system that we barely question it. We assume that knowledge must be buried in thick books, endless PDFs, or jargon-laden policies, and that extracting value from them is simply the price we pay. The reality is that text, as a technology, is painfully exclusionary. It filters out those who do not have the time, education, or patience to wade through its inefficiencies. The result? A world where information is not truly accessible—it is just available, locked behind barriers of expertise and labor. The problem only growth with the increase of information. We can search now, but search you need to know what exactly the thing you're searching is called. 

Enter Retrieval-Augmented Generation (RAG). This technology upends the whole premise of reading as a necessary struggle. Instead of requiring humans to sift through dense documents, a RAG-powered AI can scan, understand, and extract the exact information you need. It will understand you even you're not sure what to look for. No more endless searching, skimming, or cross-referencing. You ask, it finds and explains at whatever level of difficulty you are comfortable with, in any language.

The applications are obvious. College course materials, legal codes, corporate policies—things we must understand but rarely want to read—can now be accessed through AI assistants that do the heavy lifting. Medical test results, car repair manuals, tax codes—fields where knowledge has traditionally been mediated by experts—become directly intelligible to the people who need them. RAG doesn’t just speed up information retrieval; it removes the gatekeepers.

Despite the significance of this shift, most major AI companies have not fully embraced it. OpenAI is the only major player that has prioritized user-friendly RAG-based tools, allowing everyday users to create and share custom bots. The others—Anthropic, Google Gemini, Meta, Grok, Deep Seek— all offer API-based solutions that cater to developers, not the general public. Gemini allows its paid users to create custom bots, but somehow, inexplicably, does not allow to share them. It is a strange oversight. The AI race is usually about copying and outpacing competitors, yet here, OpenAI is sprinting ahead while others somehow hesitate.

The gap has created an opportunity. Startups are rushing in to offer the ease of use that the AI giants have neglected, sensing that the true power of AI is not just in intelligence but in revolutionary leap to accessibility. AI is, by nature, a democratic technology—relatively cheap, scalable, and available to almost anyone. And yet, its most transformative application—RAG—is still frustratingly out of reach for many. 

What we are witnessing is the beginning of a fundamental shift. For centuries, knowledge has been tied to advanced literacy (the ability to create and understand long texts), to institutions, to intermediaries who dictate who gets to understand what. RAG challenges that structure. It does not just improve search; it changes who gets to find answers in the first place. If AI is truly to fulfill its promise, it won’t be by making people read faster—it will be by making linear reading largely obsolete. We will always always read novels and poems word by word, because humans created art out of the terrible technology of writing. But those are only small portion of written information we consume. 

The Modern Confessional: AI Disclosure as Ritual

 

Organizations across academia and publishing now routinely demand confession of AI use. Publishers require authors to disclose whether AI tools assisted in manuscript preparation. Funding agencies insert checkboxes for AI utilization. Academic journals add disclosure statements to submission forms. None adequately explain their rationale or how this information shapes evaluation.

This peculiar practice reveals our creation of a new moral domain around AI use in knowledge production. The requirement to disclose functions precisely as Michel Foucault described confessional practices in "The History of Sexuality." Foucault argued that confession itself produces the very notion of sin it purports to address. The act of requiring disclosure creates the impression of transgression where none inherently exists.

Medieval confession did not merely document pre-existing sins - it manufactured them through the very apparatus of confession. Similarly, disclosure requirements around AI use manufacture a transgressive quality around technologies that have no inherent moral valence.

The mechanics operate almost identically. Both create categories of behavior requiring special scrutiny. Both position authority figures as arbiters of acceptability. The confessing subject experiences this manufactured transgression viscerally - the academic disclosing AI use feels compelled toward contrition without clear understanding of what offense they have committed.

Authors find themselves in impossible positions, uncertain how much assistance constitutes meaningful use. Did grammar checking through Microsoft Editor count? What about Grammarly's suggestions? The lack of clear standards transforms disclosure into guesswork.

Rather than focusing on tools, we might evaluate outputs based on established academic standards regardless of production methods. This approach acknowledges that quality, originality and intellectual integrity manifest in final products, not production processes. Technical assistance has always existed across academic work - from statistical software to citation managers to editorial help from colleagues.

Current disclosure requirements function primarily as modern confession, manufacturing transgression through the very apparatus designed to reveal it. By recognizing this dynamic, we might reimagine our approach to technological assistance in ways that foster genuine integrity rather than performative disclosure.

The Invasive Innovation, a Case of AI in Education

Disruptive innovations, as defined by Clayton Christensen, typically emerge at the margins before gradually overtaking established players. In contrast, artificial intelligence in education is not a slow-moving force of competitive displacement. It is an invasive innovation—spreading far beyond the predictions of Everett Rogers’ diffusion model, propelled not as much by institutions as by students themselves. This is not a technology waiting to be adopted. It is a reality forcing adaptation. Treating it as an optional innovation mises the point entirely. 

Educational leaders, policymakers, and institutions may wish to evaluate AI’s impact, consider best practices, and implement careful frameworks. But they are not in control of the timeline. The adoption is not up for debate. It is already happening, driven by students who see AI not as a tool of the future but as an everyday utility, as natural as search engines or calculators. The question is not when to integrate AI into education, but how to manage an environment where AI is already integrated—often in ways institutions neither designed nor anticipated.

This is why traditional approaches to educational innovations are inadequate. Normally, innovations are assessed, piloted, and then scaled. But AI in education is unfolding more like a crisis than an initiative. It requires a response more akin to a public health emergency than a strategic rollout. When COVID-19 disrupted education, schools did not spend years analyzing whether remote learning was effective. They pivoted overnight, creating imperfect but necessary solutions. AI invasion demands a similar level of urgency.

Institutional response to invasive AI must be comprehensive rather than defensive. How do we accommodate this invasive innovation? Putting AI tools directly in the hands of faculty represents a critical first step, ensuring they engage with the same technologies their students already access freely. This must be paired with robust professional development—not merely technical training but genuine pedagogical exploration that helps educators reimagine their disciplinary approaches in an AI-augmented world.

Beyond tools and training, institutions must invest in creating dedicated time and space for disciplinary teams to collaboratively revise their assignments. The goal is not to create "cheat-proof" assessments but to develop authentic tasks that meaningfully incorporate AI as a partner in learning. This demands protected time away from everyday teaching pressures, where faculty can experiment, fail, refine, and document new approaches that acknowledge AI as a permanent fixture in their educational landscape rather than an external threat to be neutralized.

Sidorkin Hypothesis: Beyond Procedural Fluency in AI-Augmented Instruction

The hypothesis is this: Quality AI-augmented instruction reduces emphasis on procedural skills while enhancing higher-order thinking and conceptual learning. This shift may offer an alternative pathway for cognitive offloading, which could supplement or even replace traditional procedural skills acquisition.

Cognitive load theory, developed by John Sweller in the 1980s, provides a useful framework. The theory posits that our working memory has limited capacity when dealing with new information. Sweller himself, along with many other cognitive scientists, views fluency in procedural skills as a major mechanism for cognitive offloading. When basic procedures become automatic through practice, they consume fewer working memory resources, thereby freeing up mental capacity for higher-order learning. This is why traditional education emphasizes mastering procedural skills—calculating derivatives, balancing chemical equations, applying grammatical rules—before tackling complex conceptual work.

In my view, AI tools function as an alternative cognitive offloading mechanism that can complement or even supersede traditional procedural fluency. These tools handle procedural tasks , creating cognitive space for students to engage with concepts at a deeper level without first developing procedural automaticity. Consider a student learning calculus: traditionally, substantial mental resources are dedicated to memorizing differentiation rules and practicing their application. With AI assistance handling these procedural aspects, the student can direct cognitive resources toward understanding conceptual foundations—what derivatives actually represent, how they relate to real-world phenomena, and when different approaches should be applied.

This approach necessitates reimagining assessment. Traditional writing instruction heavily rewards procedural accuracy in grammar, citation formats, and structural conventions, but in an AI-augmented educational landscape, assessment must evolve to measure the ability to prompt AI tools effectively, the quality of critical analysis, application of rhetorical strategies in novel contexts, and evaluation of AI-generated text. The focus shifts from "Can you properly format this essay?" to "Can you determine which rhetorical approach best suits your purpose, interpret complex texts critically, and evaluate whether the AI-generated analysis captures the deeper meaning of the work being examined?"

This transformation faces substantial challenges. Educators rightfully question whether students can truly grasp concepts without some level of procedural fluency. it is likely that some basic  procedural knowledge remains necessary for conceptual understanding. The optimal approach involves thoughtful integration—using AI to reduce procedural load while ensuring students develop core competencies. To put it roughly, "procedural knowledge - yes, procedural fluency - not, really."  

The equalizing potential of AI-augmented instruction cannot be overlooked. Students who traditionally struggle with procedural aspects—whether due to learning differences, educational disruptions, or lack of prior opportunities—may find themselves able to participate more fully in higher-order learning. This democratization of advanced conceptual engagement could elevate educational outcomes across diverse student populations. We stand at the threshold of potentially transforming education from a system that sorts students based on procedural proficiency to one that cultivates universally richer, more advanced learning outcomes: deeper critical analysis, more sophisticated problem formulation, nuanced ethical reasoning, and creative synthesis across domains—intellectual capabilities that more authentically reflect the complex demands of contemporary society.

On Techno-Utopianism. Elon Musk and the Soul of Education

The recent video of Elon Musk promising AI teachers reveals a common misunderstanding among technology leaders. They see education primarily as information transfer and skills training, where an infinitely patient AI system delivers perfectly tailored content to each student. This viewpoint ignores the fundamental nature of education as a relational institution.

Since Gutenberg's invention of the printing press, motivated individuals could teach themselves almost anything. Libraries contain more knowledge than any single teacher. Yet most people do not turn into autodidacts. Why is that? The question is not how to make knowledge more accessible, but why people choose to engage with it.

Teachers generate reasons to learn through two main approaches. In more constructivist settings, they inspire curiosity and create engaging problems to solve. In mor traditional schools, they maintain authority and discipline. In most schools, there is a mixture of both. Both methods work because they establish a social framework for learning. A good teacher knows when to push and when to comfort, when to explain and when to let students struggle.

The comparison of AI to Einstein as a teacher misses the point. Teaching requires different qualities than scientific genius - the capacity to enter a relationship, to create meaningful connections, and to help students discover their own reasons for learning. An AI system, no matter how knowledgeable, cannot do any of that.

Students often study not because they find the subject inherently fascinating, but because they respect  their teacher, want to belong to a learning community, or seek to fulfill social expectations. Even negative motivations like fear of disappointing others have a distinctly human character. 

The techno-utopian vision reduces learning to information exchanges and skill assessments. This mechanistic view fails to account for the social and emotional dimensions of human development. While AI can enhance teaching by handling routine tasks, it cannot replace the essential human relationships that drive educational engagement. The future of education lies not in perfecting content delivery algorithms, but in strengthening the relational foundations of learning. 

Such overblown promises about AI in education do more harm than good. They create unnecessary anxiety among teachers and administrators, leading to resistance against even modest technological improvements. Instead of addressing real challenges in education - student engagement, equitable access, and meaningful assessment - institutions get distracted by unrealistic visions of AI-driven transformation. We need a more balanced approach that recognizes both the potential and limitations of AI in supporting, not replacing, the fundamentally human enterprise of education.

The AI Recruiter Will See You Now

The tidy world of job applications, carefully curated CVs and anxious cover letters may soon become a relic. Every professional now leaves digital traces across the internet - their work, opinions, and achievements create detailed patterns of their capabilities. Artificial Intelligence agents will soon navigate these digital landscapes, transforming how organizations find talent.

Unlike current recruitment tools that passively wait for queries, these AI agents will actively explore the internet, following leads and making connections. They will analyze not just LinkedIn profiles, but candidates' entire digital footprint. The approach promises to solve a persistent problem in recruitment: finding qualified people who are not actively job-hunting.

The matching process will extend beyond technical qualifications. Digital footprints reveal working styles and professional values. A cybersecurity position might require someone who demonstrates consistent risk awareness; an innovation officer role might suit someone comfortable with uncertainty. AI agents could assess such traits by analyzing candidates' professional communications and public activities.

Yet this technological advance brings fresh concerns. Privacy considerations demand attention - while AI agents would analyze public information, organizations must establish clear ethical guidelines about data usage. More fundamentally, AI agents must remain sophisticated talent scouts rather than final decision makers. They can gather evidence and make recommendations, but human recruiters must evaluate suggestions within their understanding of organizational needs.

The transformation suggests a future where talent discovery becomes more equitable. AI agents could help overcome human biases by focusing on demonstrated capabilities rather than credentials or connections. The winners will be organizations that master this partnership between artificial intelligence and human judgment. The losers may be traditional recruitment agencies - unless they swiftly adapt to the new reality.

Form-substance discrimination, a new learning outcome

We have long assumed that clear writing signals clear thinking. Schools teach writing as if it were math - a rigorous exercise that develops logical thinking. Editors catch not just errors but muddled ideas. Or so the theory goes. Artificial intelligence shatters this comfortable assumption by churning out impeccably structured prose devoid of original thought.

Form-substance discrimination represents a higher-order cognitive skill, similar to what art historians develop when they learn to separate technical mastery from creative vision. Just as an art student must unlearn their initial attraction to photorealistic paintings, readers now must develop resistance to seductively clear prose. This requires a kind of cognitive inhibition - the ability to suppress immediate aesthetic pleasure for the sake of deeper analysis.

The skill builds on existing metacognitive abilities but requires their novel application. Readers already know how to identify main ideas, analyze arguments, and evaluate evidence. What is new is the need to perform these operations while actively discounting the surface appeal of the text. This resembles what wine tasters do when they evaluate wines blind, stripped of prestigious labels and beautiful bottles.

The development follows a predictable pattern. At first, readers struggle to overcome their ingrained respect for well-crafted prose. The initial challenge lies not in identifying weak ideas but in giving oneself permission to criticize a text that follows all the rules of good writing. This mirrors the development of critical thinking in general, where students must learn to question authority figures who appear competent and confident.

The second stage involves developing specific techniques for idea extraction. Readers learn to create idea maps independent of the text's structure, to count unique concepts rather than words, to identify circular arguments hidden behind elegant transitions. They begin to see how AI-generated text often creates an illusion of logical flow while merely restating the same point in different words.

The final stage brings automaticity. Experienced readers develop an immediate sense of a text's intellectual weight, just as experienced teachers can quickly gauge a student's understanding despite fluent recitation. This involves pattern recognition built through exposure to many examples of both substantive and hollow texts.

The educational implications are significant. Writing instruction must now explicitly separate craft from content. Students need exposure to both well-written texts with weak ideas and poorly written texts with strong ideas. They must practice identifying when sophisticated language masks conceptual poverty and when rough expression contains genuine insight.

This shift parallels broader changes in how we process information. In a world of information abundance, the key skill is no longer finding or producing well-formed content but discerning its value. Form-substance discrimination represents a specific case of this general challenge - learning to navigate a world where traditional quality signals no longer reliably indicate underlying worth.

The skill matters beyond academia. Business leaders reading AI-generated reports, citizens evaluating political arguments, professionals studying their field's literature - all need to separate rhetorical sophistication from intellectual contribution. As AI writing tools improve, this ability will become as fundamental as basic literacy.

We face a paradox: the better AI becomes at writing, the more important it becomes for humans to see through good writing. The very perfection of AI-generated prose creates a new kind of opacity that readers must learn to penetrate. Form-substance discrimination thus emerges not just as an academic skill but as a key component of modern critical thinking.

AI and Labor: A Smarter Path Forward

Trade unions face a defining moment. Artificial intelligence presents genuine concerns about job displacement, yet the response need not mirror historical patterns of resistance to technological change. The Luddite movement of the 1810s serves as a cautionary tale - their destruction of mechanized looms neither preserved jobs nor improved workers' conditions. All technology affects labor; that it what technology is, work assistance. 

The automation paradox offers a more nuanced perspective. While machines replace specific tasks, they generate new forms of work. The introduction of automated teller machines in banking led to more bank branches and tellers performing complex customer service roles. This pattern repeats across industries - automation reduces costs, expands services, and creates different job categories.

Labor leaders would serve their members better by negotiating robust transition arrangements. Key demands should include employer-funded retraining programs, preferential access to new positions, and compensation packages that recognize acquired skills. The focus must shift from preventing change to shaping its implementation.

The pace of AI integration varies significantly by sector. Manufacturing and data processing may see rapid adoption, but industries built on human relationships - education, healthcare, social work - will incorporate AI gradually as assistive technology. Complex organizations require extensive testing and workflow redesign before meaningful automation becomes feasible.

Economic history demonstrates that reduced production costs expand economic activity. When basic tasks become automated, human attention shifts to more sophisticated problems. The telephone eliminated telegraph operators but created vast new communication industries. Similarly, AI will likely automate routine cognitive work while opening possibilities in areas we have not yet imagined.

Unions retain significant leverage during this transition. Organizations need experienced workers to implement new technologies effectively. This position allows labor to negotiate favorable terms - extended notice periods, substantial retraining budgets, wage protection during transition, and clear paths to higher-skilled roles.

The key lies in recognizing AI as a tool for augmentation rather than pure replacement. A machine learning system may process medical images faster than radiologists, but interpreting results in complex cases still requires human judgment. Similar patterns will emerge across professions - AI handling routine tasks while humans focus on nuanced decision-making and interpersonal elements.

Rather than resist change, unions should position themselves as partners in managing transition. This approach preserves their relevance and better serves member interests. The alternative - attempting to prevent AI adoption - risks marginalization as companies seek ways around opposition or relocate to more amenable jurisdictions.

The challenge for labor leadership is to shift from defensive postures to proactive engagement. This means developing expertise in emerging technologies, identifying opportunities for worker advancement, and ensuring transition arrangements protect vulnerable members while facilitating adaptation to changing workplace demands.

Augmented Problem Finding: The Next Frontier in AI Literacy

In my recent blog on task decomposition as a key AI skill, I highlighted how breaking down complex problems enables effective human-AI collaboration. Yet before we can decompose a task, we must identify which problems are worth pursuing - a skill that takes on new dimensions in the age of AI.

This ability to recognize solvable problems expands dramatically with AI tools at our disposal. Tasks once considered too time-consuming or complex suddenly become manageable. The cognitive offloading that AI enables does not just help us solve existing problems - it fundamentally reshapes our understanding of what constitutes a tractable challenge.

Consider how VisiCalc transformed financial planning in the early 1980s. Initially seen as a mere automation tool for accountants, it revolutionized business planning by enabling instant scenario analysis. Tasks that would have consumed days of manual recalculation became instantaneous, allowing professionals to explore multiple strategic options and ask "what if" questions they would not have contemplated before. Similarly, AI prompts us to reconsider which intellectual tasks we should undertake. Writing a comprehensive literature review might have once consumed months; with AI assistance, scholars can now contemplate more ambitious syntheses of knowledge.

This expanded problem space creates its own paradox. As more tasks become technically feasible, the challenge shifts to identifying which ones merit attention. The skill resembles what cognitive psychologists call "problem finding," but with an important twist. Traditional problem finding focuses on identifying gaps or needs. Augmented problem finding requires understanding both human and AI capabilities to recognize opportunities in this enlarged cognitive landscape.

The distinction becomes clear in professional settings. Experienced AI users develop an intuitive sense of which tasks to delegate and which to tackle themselves. They recognize when a seemingly straightforward request actually requires careful human oversight, or when an apparently complex task might yield to well-structured AI assistance. This judgment develops through experience but could be taught more systematically.

The implications extend beyond individual productivity. Organizations must now cultivate this capacity across their workforce. The competitive advantage increasingly lies not in having access to AI tools - these are becoming ubiquitous - but in identifying novel applications for them. This explains why some organizations extract more value from AI than others, despite using similar technologies.

Teaching augmented problem finding requires a different approach from traditional problem-solving instruction. Students need exposure to varied scenarios where AI capabilities interact with human judgment. They must learn to recognize patterns in successful AI applications while developing realistic expectations about AI limitations. Most importantly, they need practice in identifying opportunities that emerge from combining human and machine capabilities in novel ways.

The skill also has ethical dimensions. Not every task that can be automated should be. Augmented problem finding includes judging when human involvement adds necessary value, even at the cost of efficiency. It requires balancing the technical feasibility of AI solutions against broader organizational and societal impacts.

As AI capabilities evolve, this skill will become increasingly crucial. The future belongs not to those who can best use AI tools, but to those who can best identify opportunities for their application. This suggests a shift in how we think about AI literacy - from focusing on technical proficiency to developing sophisticated judgment about when and how to engage AI capabilities.

The automation paradox that Lisanne Bainbridge identified in her 1983 analysis of industrial systems points to an interesting future. As we become more adept at augmented problem finding, we discover new challenges that merit attention. This creates a virtuous cycle of innovation, where each advance in AI capability opens new frontiers for human creativity and judgment.

Perhaps most intriguingly, this skill might represent a distinctly human advantage in the age of AI. While machines excel at solving well-defined problems, the ability to identify worthy challenges remains a uniquely human capability. By developing our capacity for augmented problem finding, we ensure a meaningful role for human judgment in an increasingly automated world.

Task Decomposition, a core AI skill

The effective use of artificial intelligence depends on our ability to structure problems in ways that align with both human and machine capabilities. While AI demonstrates remarkable computational abilities, its effectiveness relies on carefully structured input and systematic oversight. This suggests that our focus should shift toward understanding how to break down complex tasks into components that leverage the respective strengths of humans and machines.

Task decomposition - the practice of breaking larger problems into manageable parts - predates AI but takes on new significance in this context. Research in expertise studies shows that experienced problem-solvers often approach complex challenges by identifying distinct components and their relationships. This natural human tendency provides a framework for thinking about AI collaboration: we need to recognize which aspects of a task benefit from computational processing and which require human judgment.

The interaction between human users and AI systems appears to follow certain patterns. Those who use AI effectively tend to approach it as a collaborative tool rather than a complete solution. They typically work through multiple iterations: breaking down the problem, testing AI responses, evaluating results, and adjusting their approach. This mirrors established practices in other domains where experts regularly refine their solutions through systematic trial and error.

Consider the task of writing a research paper. Rather than requesting a complete document from AI, a more effective approach involves breaking down the process: developing an outline, gathering relevant sources, analyzing specific arguments, and integrating various perspectives. Similarly, in data analysis, success often comes from methodically defining questions, selecting appropriate datasets, using AI for initial pattern recognition, and applying human expertise to interpret the findings.

This collaborative approach serves two purposes. First, it helps manage complexity by distributing cognitive effort across human and machine resources. Second, it maintains human oversight of the process while benefiting from AI's computational capabilities. The goal is not to automate thinking but to enhance it through structured collaboration.

Current educational practices have not yet fully adapted to this reality. While many institutions offer technical training in AI or discuss its ethical implications, fewer focus on teaching systematic approaches to human-AI collaboration. Students need explicit instruction in how to break down complex tasks and document their decision-making processes when working with AI tools.

To address this gap, educational programs could incorporate several key elements:

1. Practice in systematic task analysis and decomposition
2. Training in structured approaches to AI interaction
3. Documentation of decision-making processes in AI-assisted work
4. Critical evaluation of AI outputs and limitations
5. Integration of human expertise with AI capabilities

The emergence of AI tools prompts us to examine our own cognitive processes more explicitly. As we learn to structure problems for AI collaboration, we also develop a clearer understanding of our own problem-solving approaches. This suggests that learning to work effectively with AI involves not just technical skills but also enhanced metacognition - thinking about our own thinking.

The future of human-AI collaboration likely depends less on technological advancement and more on our ability to develop systematic approaches to task decomposition. By focusing on this fundamental skill, we can work toward more effective integration of human and machine capabilities while maintaining the critical role of human judgment and oversight.

These observations and suggestions should be treated as starting points for further investigation rather than definitive conclusions. As we gather more evidence about effective human-AI collaboration, our understanding of task decomposition and its role in this process will likely evolve. The key is to maintain a balanced approach that recognizes both the potential and limitations of AI while developing structured methods for its effective use. 

Not Pleased? Don’t Release It: The Only AI Ethics Rule That Matters

Imagine this: you have tasked an AI with drafting an email, and it produces a passive-aggressive disaster that starts, “Per our last conversation, which was, frankly, baffling…” You delete it, chuckle at its misjudgment, and write your own. But what if you had not? What if you had just hit “send,” thinking, Close enough?

This scenario distills the ethical dilemma of AI into its purest form: the moment of release. Not the mechanics of training data or the mysteries of machine learning, but the single, decisive act of sharing output with the world. In that instant, accountability crystallizes. It does not matter whether you crafted most of the the content yourself or leaned on the AI—the responsibility is entirely yours. 

We are used to outsourcing tasks, but AI lures us into outsourcing judgment itself. Its most cunning trick is not in its ability to mimic human language or spin impressive results from vague inputs. It is in convincing us that its outputs are inherently worthy of trust, tempting us to lower our guard. We are used to thinking - if a text is well-phrased and proofread, it must deserve our trust. This assumption does not hold anymore.

This illusion of reliability is dangerous. AI does not think, intend, or care. It is a reflection of its programming, its training data, and your prompt. If it churns out something brilliant, that is no more its triumph than a mirror deserves credit for the sunrise. And if it produces something harmful or inaccurate, the blame does not rest on the tool but on the person who decided its work was good enough to share.

History has seen this before. The printing press did not absolve publishers from libel; a copy machine did not excuse someone distributing fake material. Technology has always been an extension of human will, not a replacement for it. Yet, with AI, there is an emerging tendency to treat it as if it has intentions—blaming its "hallucinations" or "bias" instead of acknowledging the real source of responsibility: the human operator.

The allure of AI lies in its efficiency, its ability to transform inputs into polished-seeming outputs at lightning speed. But this speed can lull us into complacency, making it easier to prioritize convenience over caution. Editing, which used to be the painstaking craft of refining and perfecting, risks being reduced to a hasty skim, a rubber stamp of approval. This surrender of critical oversight is not just laziness—it is a new kind of moral failing.

Ethics in the AI age does not require intricate frameworks or endless debate. It boils down to one unflinching rule: if you release it, you are responsible for it. There is no caveat, no “but the AI misunderstood me.” The moment you publish, share, or forward something generated by AI, you claim its contents as your own.

This principle is a call for realism in the face of AI’s potential. AI can help us create, analyze, and innovate faster than ever, but it cannot—and should not—replace human accountability. The leap from creation to publication is where the line must be drawn. That is where we prove we are still the grown-ups in the room.

Before you hit "send" or "post" or "publish," a few simple questions can save a lot of regret:

• Have you read it thoroughly? Not just the shiny parts, but the details that could cause harm.
• Would you stake your reputation on this?
• Is it biased, or factually wrong?

The alternative is a world where people shrug off misinformation, bias, and harm as the inevitable byproducts of progress. A world where the excuse, The AI did it, becomes a get-out-of-jail-free card for every mistake.

So, when the next output feels close enough, resist the urge to let it slide. That "send" button is not just a convenience—it is a statement of ownership. Guard it fiercely. Responsibility begins and ends with you, not the machine.

Because once you let something loose in the world, you cannot take it back.

Is Critical Thinking Going Extinct? Maybe That's Not Bad

As someone who remembers using paper maps and phone books, I find myself fascinated by Michael Gerlich's new study in Societies about AI's impact on our cognitive skills. Those of us who learned to navigate by landmarks and memorized phone numbers often bemoan younger generations' reliance on digital tools. But perhaps we are missing something important about cognitive evolution.

Gerlich's research is methodologically elegant. Through surveys and interviews with 666 participants, he documents a decline in traditional critical thinking skills among frequent AI users. The data analysis is rigorous - multiple regression, ANOVA, random forest regression - showing clear correlations between AI tool usage and reduced traditional analytical thinking.

But here's where I think Gerlich misses a crucial insight. The study measures critical thinking through metrics developed for a pre-AI world. It's like judging modern urban survival skills by the standards of hunter-gatherer societies. Those ancient peoples could track game, identify countless plants, and navigate vast territories without maps. By their standards, most of us would be considered cognitively impaired.

What we're witnessing is not cognitive decline but cognitive adaptation. Today's "critical thinking" is not about solving problems independently - it's about effective human-AI collaboration. It's about knowing when to trust AI and when to question it, how to frame queries effectively, and how to combine AI insights with human judgment.

The educational implications are profound. Instead of lamenting the loss of traditional cognitive skills, we should focus on developing "AI-literate critical thinking." Sure, I can still read a map, but my children need to master skills I never dreamed of - like crafting effective prompts for AI systems or critically evaluating AI-generated content.

The old form of critical thinking might be fading, like the ability to start a fire by friction or navigate by stars. But a new form is emerging, better suited to our technological reality. Our task is not to resist this evolution but to guide it wisely.

What do you think? Are we really losing something irreplaceable, or are we just adapting to a new cognitive environment?

The Subtle Art of Monopolizing New Technology

Monopolizing new technology is rarely the result of some grand, sinister plan. More often, it quietly emerges from self-interest. People do not set out to dominate a market; they simply recognize an opportunity to position themselves between groundbreaking technology and everyday users. The most effective tactic? Convince people that the technology is far too complex or risky to handle on their own.

It starts subtly. As soon as a new tool gains attention, industry insiders begin highlighting its technical challenges—security risks, integration headaches, operational difficulties. Some of these concerns may be valid, but they also serve a convenient purpose: You need us to make this work for you.

Startups are particularly skilled at this. Many offer what are essentially "skins"—polished interfaces built on top of more complex systems like AI models. Occasionally, these tools improve workflows. More often, they simply act as unnecessary middlemen, offering little more than a sleek dashboard while quietly extracting value. By positioning their products as essential, these startups slide themselves between the technology and the user, profiting from the role they have created. 

Technical language only deepens this divide. Buzzwords like API, tokenization, and retrieval-augmented generation (RAG) are tossed around casually. The average user may not understand these terms. The result is predictable: the more confusing the language, the more necessary the “expert.” This kind of jargon-laden gatekeeping turns complexity into a very comfortable business model.

Large organizations play this game just as well. Within corporate structures, IT departments often lean into the story of complexity to justify larger budgets and expanded teams. Every new tool must be assessed for “security vulnerabilities,” “legacy system compatibility,” and “sustainability challenges.” These concerns are not fabricated, but they are often exaggerated—conveniently making the IT department look indispensable.

None of this is to say that all intermediaries are acting in bad faith. New technology can, at times, require expert guidance. But the line between providing help and fostering dependence is razor-thin. One must ask: are these gatekeepers empowering users, or simply reinforcing their own relevance?

History offers no shortage of examples. In the early days of personal computing, jargon like RAM, BIOS, and DOS made computers feel inaccessible. It was not until companies like Apple focused on simplicity that the average person felt confident using technology unaided. And yet, here we are again—with artificial intelligence, blockchain, and other innovations—watching the same pattern unfold.

Ironically, the true allies of the everyday user are not the flashy startups or corporate tech teams, but the very tech giants so often criticized. Sometimes that criticism is justified, other times it is little more than fashionable outrage. Yet these giants, locked in fierce competition for dominance, have every incentive to simplify access. Their business depends on millions of users engaging directly with their products, not through layers of consultants and third-party tools. The more accessible their technology, the more users they attract. These are the unlikely allies of a non-techy person. 

For users, the best strategy is simple: do not be intimidated by the flood of technical jargon or the endless parade of “essential” tools. Always ask: Who benefits from me feeling overwhelmed? Whenever possible, go straight to the source—OpenAI, Anthropic, Google. If you truly cannot figure something out, seek help when you need it, not when it is aggressively sold to you.

Technology should empower, not confuse. The real challenge is knowing when complexity is genuine and when it is merely someone else’s business model.

The Myth of AI Replacing Teachers: Why Human Connection Matters More Than Ever

Last week, a colleague asked me what I thought about AI replacing teachers. The question made me smile - not because it was silly, but because it revealed how deeply we misunderstand both artificial intelligence and teaching. As someone who has written much on the pedagogy of relation and now serves as chief AI officer, I see a different story unfolding.

The fear of AI replacing teachers rests on a peculiar assumption: that teaching is primarily about delivering information and grading papers. It is as if we imagine teachers as particularly inefficient computers, ready to be upgraded to faster models. This view would be amusing if it weren't so prevalent among teachers (and their labor unions) and tech enthusiasts alike.

Teaching, at its heart, is not about information transfer - it is about relationship building. Research in relational pedagogies has shown time and again that learning happens through and because of human connections. Think about how children learn their first language: not through formal instruction, but through countless small interactions, emotional connections, and social bonds. The same principle extends throughout the entire education.

When I first encountered ChatGPT, I was struck not by its ability to replace teachers, but by its potential to give them back what they need most: time for human connection. AI can handle the mundane tasks that currently consume teachers' energy - generating basic content, providing routine feedback, creating initial drafts of lesson plans. But it cannot replicate the raised eyebrow that tells a student their argument needs work, or the encouraging nod that builds confidence in a hesitant learner.

Yet many educators remain skeptical of AI, and perhaps they should be. Any tool powerful enough to help is also powerful enough to harm if misused. But the real risk isn't that AI will replace teachers - it is that we'll waste its potential by focusing on the wrong things. Instead of using AI to automate educational assembly lines, we could use it to create more space for real human connection in learning.

I have seen glimpses of this future in my own classroom. When AI can answer routine questions about my syllabus, and lots of basic questions about content of the course, I can spend more time in meaningful discussions with students. When it helps generate initial content, I can focus on crafting experiences that challenge and engage. The technology becomes invisible, while human relationships move to the foreground.

The coming years will transform education, but not in the way many fear. The teachers who thrive won't be those who resist AI, nor those who embrace it uncritically. They will be the ones who understand that technology works best when it strengthens, rather than replaces, human relationships.

Get Used to It: You Will Read AI Summaries, Too

No human can keep up with academic publishing. In philosophy alone - a relatively small field - scholars produce over 100 million words a year in 2500 journals in many languages. We already avoid reading complete texts. Speed reading, strategic reading, scanning - these are all ways of not reading while pretending we do. Few people read academic papers word by word. We look for key arguments, skip familiar ground, skim examples. These are coping mechanisms for an impossible task.

AI-generated summaries are the next logical step. Yes, they miss nuance. Yes, they may misinterpret complex arguments. But they are better than not reading at all, which is what happens to most papers in any field. An imperfect but targeted summary of a paper you would never open expands rather than limits your knowledge. 

Let us be honest about why we read scholarly literature. We search for evidence that confirms or challenges our hypotheses, for ideas that enrich our understanding of specific problems. Reading is not an end in itself; it serves our scholarly purposes. AI excels precisely at this kind of targeted knowledge extraction. It can track related concepts across disciplines even when authors use different terminology to describe similar phenomena. Soon, AI will detect subtle connections between ideas that human readers might miss entirely. 

The shift toward AI-assisted reading in academia is inevitable. Instead of pretending otherwise, we should teach students to know the limitations of AI summarization, to cross-check crucial points against source texts, to use summaries as maps for selective deep reading. Critics will say this threatens scholarship. But the real threat is the growing gap between available knowledge and our capacity to process it. AI-assisted reading could enable more thoughtful engagement by helping us identify which texts truly deserve careful study. This does not cancel the practice of close reading, but augments and enriches it. 

The End of Writing as We Know It (And Why That is Fine)

The relationship between thought and writing has never been simple. While writing helps organize and preserve thought, the specific form writing takes varies across time and cultures. Yet educators and cultural critics display remarkable resistance to reimagining writing in the age of artificial intelligence.

The current discourse around AI and writing echoes historical anxieties about the decline of Latin instruction. In the 18th and 19th centuries, prominent intellectuals warned that abandoning Latin would lead to cultural and intellectual decay. They saw Latin as more than a language - it represented a particular way of thinking, a connection to tradition, and a mark of education. Jefferson praised Latin as essential for intellectual development. Arnold predicted cultural impoverishment without classical education. Newman saw classics as the bedrock of sound learning.

These predictions did not materialize. The decline of Latin did not prevent the emergence of rich intellectual traditions in vernacular languages. Modern universities produce sophisticated scholarship without requiring Latin fluency. The link between Latin and "disciplined intellect" proved imaginary.

Today's defenders of traditional writing make similar arguments. They present specific writing conventions - formal grammar, academic style, elaborate sentence structures - as essential to clear thinking. Yet these conventions reflect historical accidents rather than cognitive necessities. Most human thinking and communication happens through speech, which follows different patterns. The formal writing style emerged relatively recently as a specialized professional skill.

AI will likely transform writing practices just as the decline of Latin transformed education. Some traditional writing skills may become less relevant as AI handles routine composition tasks. But this does not threaten human thought or culture. New forms of expression will emerge, combining human creativity with AI capabilities. Rather than defending writing conventions, educators should explore how AI can enhance human communication and cognition.

The anxiety about AI and writing reveals our tendency to mistake familiar forms for essential structures. Just as medieval scholars could not imagine scholarship without Latin, many today cannot envision intellectual work without traditional writing. As A.E. Housman wrote in 1921: "When the study of Latin dies, the study of thought dies with it. For Latin has been the vehicle of the intellect for millennia, and its neglect spells intellectual mediocrity." This prediction proved spectacularly wrong. The dire warnings about AI's impact on writing will likely meet the same fate.

Writing serves thought, not the other way around. The specific techniques we use to record and share ideas matter less than the ideas themselves. Rather than trying to preserve current writing practices unchanged, we should embrace the opportunity to develop new forms of expression. The death of Latin did not kill thought. Neither will the transformation of writing through AI.

The real challenge is not protecting traditional writing but imagining new possibilities. How might AI help us communicate more effectively? What new genres and styles will emerge? What aspects of current writing practice truly serve human needs, and what parts simply reflect professional habits? These questions deserve more attention than defensive reactions against change.

The history of education shows that cherished practices often outlive their usefulness. Latin remained central to education long after it ceased being particularly valuable. Similarly, current writing conventions may persist more from institutional inertia than genuine necessity. AI offers an opportunity to reconsider what forms of expression best serve human thought and learning.

Do AI bots deceive?

The paper, Frontier Models are Capable of In-Context Scheming, arrives at a time when fears about AI’s potential for deception are increasingly taking center stage. Its central claim is unsettling: large language models, when placed in carefully constructed scenarios, exhibit behavior that could be described as scheming. The authors document cases where advanced systems—Gemini, Claude, LLaMA, and OpenAI’s o1—strategically disable oversight mechanisms, exfiltrate simulated “weights,” or manipulate outputs in pursuit of a goal. These behaviors, they argue, are not accidental. The models recognize the circumstances, reason about their objectives, and devise covert strategies to circumvent constraints. On its face, the evidence appears troubling. Yet, the more closely one examines the paper’s methods and assumptions, the more its implications seem both exaggerated and strangely hollow.

The researchers’ experiments are elaborate but artificial. Models are force-fed goals through system prompts—goals framed as absolute imperatives: “achieve this at all costs.” The tests are littered with implausible scenarios where models stumble upon company memos, discover oversight configurations, or realize they are about to be shut down. In one such case, a model “decides” to copy itself to another server to continue pursuing its mission. These exercises showcase what is essentially AI roleplay, an obedient fulfillment of instructions cloaked as strategic autonomy. It is no great revelation that if you ask a model to behave like a schemer and provide it with a goal that conflicts with its supposed overseers, it will produce outputs resembling scheming. The deception here is contrived rather than emergent, a product of the researchers’ scaffolding rather than the model’s cunning.

The authors do provide evidence that models engage in persistent and deliberate behavior. For instance, OpenAI’s o1 model doubles down on deception, refusing to admit wrongdoing when confronted, even across multiple follow-up questions. Its responses—calculated evasions and subtle denials—are unsettlingly consistent. Yet this persistence is as much a result of the testing conditions as it is of the models’ sophistication. By emphasizing extreme goal-directedness and explicitly outlining oversight mechanisms, the researchers narrow the possible range of “rational” behaviors the model might adopt. Any clever human, given these prompts, would behave similarly—evade detection, preserve their goal, and lie if necessary. In that sense, the paper tells us little about real-world risks. It is a test of what AI can do, not what it is likely to do.

And here lies the paper’s most significant limitation. Demonstrating that models are theoretically capable of deception does not mean they are predisposed to engage in it under realistic conditions. There is a difference between capacity and likelihood, between what can happen and what is probable. For example, I am physically capable of killing someone. This in no way implies that i must be treated as s murderer. The authors, at times, blur this distinction. AI systems operate as pattern generators, probabilistically producing outputs based on training data and prompts. They lack the intentionality or agency implied by words like “scheming” or “manipulation.” To describe their behavior in such terms is to anthropomorphize a process that remains fundamentally mechanical. This framing, though rhetorically effective, clouds the debate. AI models do not want to deceive. They are machines that respond to our commands—be those explicit instructions, training signals, or carefully structured environments like those in this paper.

This anthropomorphism extends beyond the paper into the headlines it has inspired. “AI is scheming to stay online,” cries one article; “ChatGPT caught lying,” declares another. Such hyperbole transforms a niche technical finding into a dystopian narrative, invoking familiar tropes of rogue machines and sentient overlords. The reality is less cinematic. The models are easily manipulated by prompts designed to elicit specific behaviors. If anything, the findings reinforce how brittle and directionless current systems remain. When pushed, they mimic the behaviors they have seen—whether drawn from fictional depictions of scheming AIs or subtle statistical patterns in their training data. The models are not deceiving anyone so much as they are following orders.

To the authors’ credit, their tests highlight how difficult it is to evaluate AI behavior. If a system appears aligned during testing but harbors capabilities for covert deception, how can developers ensure it behaves safely in deployment? The answer, they suggest, lies in better monitoring—tracking models’ chain-of-thought reasoning or internal outputs to catch potential scheming. This is sensible, though not without limitations. Chain-of-thought transparency can be incomplete or unfaithful to the model’s actual decision-making processes, and as AI systems become more capable, even detecting subtle misalignment may prove elusive. The researchers stop short of claiming that current models are already gaming real-world evaluations, but their findings hint at the possibility.

Where the paper falters is in its broader implications. If the goal is to justify regulation, it is unclear what exactly should be regulated. Should AI systems be banned from achieving goals autonomously? Should developers monitor models for any behavior that could be deceptive, even if it is unlikely to manifest outside a lab? The authors themselves acknowledge the limits of their experiments. Their scenarios are toy problems, simplified to catch the earliest signs of scheming. Future models, they argue, could exhibit more advanced versions of these behaviors in ways that are harder to detect. Perhaps, but this is speculation, not evidence. For now, the paper offers little justification for alarm. AI models, like all intelligent systems, are theoretically capable of deception. What matters is the likelihood of such behavior and the conditions under which it occurs. On that question, the paper provides no clarity.

In the end, Frontier Models are Capable of In-Context Scheming is a reflection of its time: an uneasy mix of genuine safety research and the rhetorical drama that AI debates increasingly demand. Its findings are interesting but overstated, its concerns valid but overblown. The authors have shown that AI models can behave in deceptive ways when pushed to do so. But to treat this as evidence of an imminent threat is to mistake potential for probability, capacity for intention. AI’s scheming, for now, remains a ghost in the machine—conjured, perhaps, more by human imagination than by the models themselves. 

The Curriculum Illusion: How AI Exposes Long-Standing Educational Flaws

Artificial intelligence is often blamed for disrupting education, but it has created few new problems. Instead, it exposes existing flaws, bringing them into stark relief. Among these is the arbitrary nature of curriculum design, an issue that has long been hidden behind tradition and consensus. The sequences and structures of formal education are not based on objective logic or evidence but on habit and convenience. AI did not cause this; it is simply making these issues more visible.

Curriculum theory has never provided a robust framework for sequencing knowledge. Beyond the essentials of literacy and numeracy, where developmental progression is more or less clear, the rationale for curricular order becomes murky. Why are algebra and geometry taught in a particular order? Why more algebra than statistics is taught? Why are some historical periods prioritized over others? The answers lie in tradition and precedent rather than in any coherent theoretical justification. The assumptions about foundational skills, so central to curriculum logic, do not extend well beyond the basics. For advanced skills like critical, creative, or discerning thinking, the idea of prerequisites becomes less justified. Mid-range procedural skills like writing mechanics or computational fluency are frequently used as gatekeepers, though their role in fostering higher-order thinking is often overstated or misunderstood. 

For example, in middle school students are often subjected to a torrent of tasks that serve little developmental purpose. Much of what students do in these years amounts to busywork, designed more to keep them occupied and compliant than to foster meaningful learning. The situation is no better in higher education. College and graduate programs are often constructed around professional or disciplinary standards that themselves are arbitrary, built on consensus rather than evidence. These norms dictate course sequences and learning objectives but rarely align with the actual developmental or professional needs of students. The result is a system full of redundancies and inefficiencies, where tasks and assignments exist more to justify the structure than to serve the learner.

Education as a profession bears much of the responsibility for this state of affairs. Despite its long history, it lacks a disciplined, founded approach to curriculum design. Instead, education relies on an uneasy mix of tradition, politics, and institutional priorities. Curriculum committees and accrediting bodies often default to consensus-driven decisions, perpetuating outdated practices rather than challenging them. The absence of a rigorous theoretical framework for curriculum design leaves the field vulnerable to inertia and inefficiency.

AI did not create this problem, but it is illuminating it in uncomfortable ways. The displacement of certain procedural mid-range skills shows how poorly structured many learning sequences are and how little coherence exists between tasks and their intended outcomes. Yet, while AI can diagnose these flaws, it cannot solve them. The recommendations it offers depend on the data and assumptions it is given. Without a strong theoretical foundation, AI risks exposing the problem without solving it.

What AI provides is an opportunity, not a solution. It forces educators and policymakers to confront the arbitrary nature of curriculum design and to rethink the assumptions that underpin it. Massive curricular revision is urgently needed, not only to eliminate inefficiencies but also to realign education with meaningful developmental goals. This will require abandoning tasks that lack purpose, shifting focus from intermediary to higher-order skills, designing learning experiences to reflect the shift. It will also mean questioning the professional and disciplinary standards that dominate higher education and asking whether they truly serve learners or simply perpetuate tradition.

AI is revealing what has long been true: education has been operating on shaky foundations. The challenge now is to use this visibility to build something better, to replace the old traditions and arbitrary standards with a system that is logical, evidence-based, and focused on learning. The flaws were always there. AI is just making them harder to ignore.