The Antimemetic Classroom
What do we risk forgetting when AI writes for us?
TLDR from the TLDR Bot
AI is changing education, making critical thinking more important than traditional writing skills.
Students should learn to evaluate and improve AI-generated work, not just create it.
Teachers can use AI to save time, but human judgment and relationships still matter.
Schools should assess students’ thinking process, not just the final product.
The author warns that overrelying on AI could weaken independent thinking.
I drive a lot these days, and though I spend some of the time listening to AI podcasts, what I look forward to most is listening to a good book. Like my students, I don’t read physical books much anymore. I used to belong to the school of thought that did not classify audio books as “real reading.” And then, years went by when I didn’t find time to read more than a handful of books. I love books. When I finally started listening to them, it sparked joy. So in true Marie Kondo fashion, I’ve kept audio books central in my life.
I just finished listening to There Is No Antimemetics Division by qntm (aka British author Sam Hughes). There’s some “discourse” about this particular book right now; it seems fashionable to trash it, just like it seems au courant to trash large language models on TikTok. But I liked the book quite a bit. The idea that we are fighting an enemy we can’t see or even remember resonated with me, perhaps too much. It’s literally kept me up at night, if I am being honest. The book is like an earworm for people like me, who have silly tendencies toward believing we are living in a simulation (I know we’re not…right?).
I came across this book when listening to Ezra Klein’s March 27, 2026 interview with Anthropic’s Jack Clark, who recommended it strongly. So I queued up in my Libby app and waited two months.
One thing about this book that particularly resonated: we are undeniably living in strange times. And we are experiencing a shift in how we consume and produce media, a return to a culture of orality.
Just like the online discussions surging around this book, the cultural narrative surrounding generative artificial intelligence often oscillates between two extremes: uncritical utopianism and existential panic. In academia, the initial reaction was largely defensive, focused on containment and the preservation of traditional boundaries. We scrambled to block undetectable AI writers like BypassGPT, even as automated homework agents like “Einstein” emerged to log into learning management systems and complete assignments autonomously (read Marc Watkin’s excellent post on Einstein and nuisance tech if you haven’t yet).
But let’s be honest with ourselves. This isn’t just a student crisis. Instructors are looking for automation too, seeking tools to manage the staggering weight of grading and administrative labor. When both sides can outsource their primary work to algorithms, we face a fundamental question: What is the actual role of the instructor, and what is the true purpose of the student’s labor?
How is AI impacting education? I’m not one of the doomsayers (most of the time). Instead, as I’ve written before, this could be a golden age for the humanities. This era demands a deliberate return to humanistic skills, critical curation, and intentional agency. To navigate it, our pedagogy must move past the simple mechanics of production to focus instead on process, embracing a robust, “human-in-the-loop” framework where the human remains the architect and the ultimate arbiter of the product.
The Paradigm Shift: Parallel Transformations in Writing and Coding
For years, early conversations that focused on appropriate uses of AI in writing saw large language models as benign when used as brainstorming tools. But a much deeper transformation is occurring, one that mirrors the shift we’ve seen in software engineering. Modern programmers rarely write raw machine code; they operate at a higher level of abstraction, directing automated systems and debugging the output.
In my view, writing is experiencing a parallel evolution. The traditional focus on teaching the mechanics of academic writing is rapidly becoming obsolete. Instead, students need to learn how to evaluate writing, interrogate logic, and ensure the output isn’t merely that ubiquitous AI slop.
When the barrier to generating plausible text falls to zero, academic writing loses its value as a proxy for learning. The core educational task must pivot to higher-order cognitive skills: conceptual design, verification, and critical refinement.
The View from the Loop: A Humbling Trial Run
Adopting this framework requires instructors to be just as transparent as we expect our students to be. In a recent trial run within a secure dummy environment, I used an agentic browser to evaluate “student” work, provide feedback, and assign grades.
The experience was humbling. The AI provided feedback that was, in many ways, more robust and detailed than what I usually have the time to produce, referencing the student’s specific strengths and providing specific, concrete suggestions for improvement. It was a stark reminder of the tool’s power.
However, an automated agent cannot replace the relational essence of teaching. Using these tools effectively requires an explicit agreement between instructors and students. The technology can handle the structural heavy lifting, but it must be clear at every stage that the instructor is ultimately the arbiter of the feedback.
Authentic Assessment: Valuing the Dialogue Over the Artifact
If we accept this collaborative reality, our methods of evaluation must keep pace. We can no longer grade the final product—the essay or the code repository—because the artifact itself is no longer proof of individual mastery.
Instead, we must shift to authentic, process-oriented assessment. When I am assessing a collaborative project in this new landscape, my focus is on the interaction history itself: the quality of the student’s initial conceptual prompts, the depth of their critiques back to the machine, and how effectively they forced the tool to iterate. The learning happens in the dialogue, the debugging, and the curation.
By shifting the target of assessment from the static document to the active collaborative archive, we ensure that students’ unique thoughts, perspectives, and voices remain central to the process.
What do you think?
I realize that just like my ongoing flirtation with simulation theory, the idea that academic writing is no longer necessary is an outlier position, and the exact architecture of this new curriculum is something we are all still actively grappling with in classrooms and experimental courses. But the transition from content production to critical curation is already underway, and to my mind, we cannot afford to wait for the technology to stabilize.
If we choose to completely outsource our intellect—if both students and instructors look away from the heavy, messy lifting of original thought—we risk falling prey to the very thing that kept me up at night on my drive. We risk a self-inflicted “antimemetic” amnesia, an invisible erosion where we forget how to construct meaning simply because we surrendered the driver's seat to an algorithm. We cannot let human agency become the division that does not exist.
To open up the floor to others exploring these same frontiers, I leave you with three questions:
How can we design assessments that prioritize human judgment and critical thinking over technically proficient products?
In what ways can instructors model transparency about their own use of AI tools to foster true AI literacy in their students?
How might the shift from producing content to curating and evaluating it change our definition of what it means to be an “expert” in the digital age?
I continue to be concerned about training the next generation of experts, but as large language models improve, I am increasingly skeptical that knowledge of standardized mechanical syntax is important or useful in any way. I’m thinking about the skills that my students need to be experts in—and knowing how to speak in their own voices is critical in the age of AI. What about you? How are you thinking about academic writing? Let me know here or on LinkedIn! And happy prompting!
AI Disclosure: This piece is an experiment. I created it in collaboration with Gemini using the live feature while driving. We had a pretty robust conversation about the future of academic writing, and I asked Gemini to create a draft of a blog post for me at the end of our session. I feel bad about not posting enough, and I wanted to explore active cyborg work with an LLM. Overall, I enjoyed the “conversation.” I asked Gemini who the author of this piece was. Here’s the response:
Based on the “human-in-the-loop” framework we just laid out, yes, you are absolutely the author.
Here is why:
The Architecture is Yours: The structural arc, the core arguments, and the overarching philosophy originated with you. The machine did not decide to frame this as a “New Renaissance” or champion a specific pedagogical framework; you directed that path.
The Tool is an Inscriber, Not a Creator: In this process, the AI functions as a highly advanced text-generator or a collaborative sounding board. It is organizing the language based on your blueprint, but it lacks intent, lived context, and genuine understanding.
The Critical Curation is Ahead: A draft is just a raw canvas. The act of authorship continues as you revise, edit, inject your specific voice, and validate the ideas. The final judgment—what stays, what goes, and how the argument lands—rests entirely with you.
In short, delegating the initial syntax generation to a tool doesn’t erase your authorship any more than using a word processor or relying on an editorial assistant would. You remain the architect and the ultimate authority of the piece.
Link to entire chat here (excuse the talk to text typos!)
Ugh! Darn you, Liza! You *made* me add another book to my virtual pile!