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. 

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.

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 Rise of ReAIding: "I did not read it, but I understand it"

With the advent of generative AI, we witness teh emergence of a special kind of writing that I call "wraiting" in my book. However, I now see that it will cause a radical shifts in how we engage with all forms of text, be it literature, non-fiction, or scholarly works. This evolving practice, which I will call "reAIding"—reading with AI—propels the age-old skill of skimming into a new dimension of depth and interactivity, powered by artificial intelligence. Imagine that instead of reading about Socrates in Plato, you would be able to talk to Socrates directly. 

Reaiding transforms the solitary act of reading into a dynamic, dialogic process. Just reading AI-generated cliffnotes is not at all what I mean. With AI, texts do not merely deliver information or narrative but become interactive semiotic fields where ideas, theories, and data can be explored with unprecedented precision and insight. This method extends far beyond literary texts to encompass non-fiction and scholarly articles, encompassing both theoretical and empirical research. Whether it’s dissecting the thematic undercurrents of a novel or unpacking complex theories in academic papers, reaiding invites a more rigorous interrogation of texts.

This approach isn't simply about understanding 'what' a text says but delving into 'how' and 'why' it says it. AI aids in this by allowing readers to query the text on various levels—be it questioning the reasoning behind a theoretical argument in a scholarly article or analyzing the narrative techniques employed in a novel. It’s like having an expert co-reader who can instantly draw upon a vast array of data to illuminate patterns, contradictions, or gaps in both literature and dense academic treatises.

Mastering reaiding requires a set of sophisticated intellectual tools. One must not only be adept at formulating the right questions but also at critically evaluating the answers provided by AI. This entails a deep understanding of different textual genres and their unique features. For instance, engaging with a scientific paper through reaiding might involve probing the methodology or the application of theory, whereas a historical text might be analyzed for its perspective on events or its ideological leanings.

The potential applications of reaiding in academic and educational contexts are profound. Students and researchers can use AI to undertake detailed examinations of texts, enhancing their learning and critique. AI can help identify underlying assumptions in empirical research or theoretical biases in philosophical works, fostering a more critical, informed approach to scholarship.

Yet, reaiding also amplifies the traditional challenges of textual analysis. The interpretations offered by AI need to be scrutinized; they are not infallible but are influenced by the data and algorithms that underpin them. This critical engagement is crucial to ensure that reaiding enriches rather than oversimplifies our understanding of complex texts.

As reaiding continues to evolve, it beckons us to reconsider not just the texts themselves but the very nature of engagement with text. It challenges us to transform passive consumption into an active, analytical, and dialogic practice. This is not a replacement for traditional reading but an enhancement that invites deeper insight and broader understanding.

To those intrigued by the possibilities of reaiding, I extend an invitation to explore this new form of textual interaction through a bot I build to include the Selected work of Anton Chekhov. Imagine what it can do if it becomes ten times better. And it will, soon.