From Vibe-Teaching to Flow-Teaching
Why AI in Education Needs Better Metaphors
The following essay is a more readable version of an academic paper I will present at the 18th annual International Conference of Education, Research and Innovation (ICERI2025) in November. It extends ideas from my previous article about vibe-coding here on The Augmented Educator, exploring what happens when this problematic metaphor migrates from software development into educational practice.
Coined by AI researcher Andrej Karpathy in February 2025, the term “vibe-coding” described a crucial element in the early adoption of generative AI among software developers. The phrase described a programming approach where developers guide AI through natural language conversation, selecting outputs based on intuitive assessment rather than traditional line-by-line code composition. Within weeks, the term had spread throughout the software development community, its casual tone perfectly matching the seemingly painless way AI could generate functional code.
The appeal of vibe-coding’s effortless promise soon reached classrooms. Teachers and students, facing their own pressures to integrate generative AI, began adopting comparable language. Terms like “vibe-teaching” and “vibe-learning” emerged to describe how teachers and students were using AI tools.1 A teacher might “vibe” their way through lesson planning by prompting ChatGPT until something usable appeared. Students might “vibe-learn” by iterating with AI until assignments felt complete. The metaphor suggested that educational AI use could similarly be approached through feel and intuition.
But here’s the issue: as I have already pointed out in a previous essay, the vibe metaphor mis-characterizes effective practice in both domains. Quality software development, like quality education, requires directed search and purposeful iteration, not casual browsing. When developers create robust applications, they engage in systematic exploration guided by clear objectives. They test hypotheses, evaluate outputs against specifications, and refine approaches based on feedback, which differs from the undirected trial and error that “vibing” implies. This becomes even more apparent in education, where effective practice demands intentional design and careful calibration of challenges to promote learning.
What “Vibe-Teaching” Actually Looks Like
Consider how vibe-teaching would manifest in practice. A pressed-for-time teacher needs materials for tomorrow’s lesson on photosynthesis. They open an AI assistant and type: “Create a lesson plan on photosynthesis for 9th graders.” The AI generates something that looks reasonable: an introduction, some activities, discussion questions, a worksheet. The teacher skims it, makes minor adjustments, and considers the task complete.
What’s missing from this interaction? The teacher hasn’t considered their specific students’ prior knowledge or misconceptions. They haven’t thought about how this lesson connects to yesterday’s discussion or next week’s lab. The AI cannot know that three students in second period have reading difficulties, or that the class became fascinated by a question about plant respiration last Tuesday. These contextual factors that expert teachers constantly consider vanish when teaching materials simply emerge from generic prompts.
Students might engage in similar practices. Facing an essay assignment, they could turn to AI seeking a finished product. “Write a five-paragraph essay about the causes of World War I.” When the result seems too simple, they refine: “Make it more sophisticated.” “Add another historical example.” They evaluate outputs through surface features rather than understanding. “Does it sound academic enough? Is it the right length?” The actual learning that should occur through researching, evaluating sources, and constructing arguments never happens.
It has to be noted that these practices exist on a spectrum. Some teachers might thoughtfully adapt AI suggestions based on pedagogical expertise. And some students could use AI for brainstorming while maintaining intellectual ownership. But the problem is that the vibe metaphor provides no framework for distinguishing beneficial from harmful uses. Its emphasis on intuitive, affect-driven interaction offers no criteria beyond subjective and instant intellectual satisfaction.
Why Vibing Fails Education
The incompatibility between vibe metaphors and educational purposes reveals itself across multiple dimensions. First, effective pedagogy begins with clear learning objectives that inform every subsequent decision. Teachers engage in what educators call “backward design,” starting with desired outcomes and working backward to create aligned experiences. This process requires deliberate planning, not intuitive browsing. When teachers vibe their way through lesson creation, they replace intentional design with chance. Even when resulting materials appear adequate, they lack the coherent progression that characterizes expert instruction.
Second, psychological research consistently shows that struggle and effort are essential components of learning. Concepts like “desirable difficulties” highlight how cognitive challenge drives retention and transfer. When students create AI-generated essays without engaging in the writing process, they bypass the necessary struggle. Writing involves analyzing sources, organizing ideas, and finding language for complex thoughts. Each activity strengthens intellectual capacity. The AI-generated essay may earn a grade, but it leaves the student’s capabilities unchanged.
Third, the vibe framework masks essential professional labor. Teaching expertise develops through years of practice and refinement. Expert teachers possess deep content knowledge and finely tuned abilities to read student understanding. When these teachers use AI tools properly, they apply this expertise in evaluating and adapting outputs. They recognize when suggested examples might confuse rather than clarify. They understand how to sequence activities for their particular students. But if teaching involves simply generating materials that feel right, what distinguishes the professional educator from anyone with AI access?
The systemic implications extend further. Educational institutions adopting vibe approaches risk creating environments where students with access to thoughtfully integrated AI receive enhanced education, while those whose teachers rely on vibing receive generic instruction. The apparent democratization of content through AI could paradoxically increase inequality. Meanwhile, routine submission of AI-generated work threatens the meaning of assessment itself. If credentials no longer reflect actual learning, the entire educational enterprise collapses into an empty performance.
An Alternative Framework: Flow Theory and Education
Rather than allowing imprecise metaphors to shape practice, we need frameworks grounded in established theory. As I have pointed out in my previous post on the problems of vibe-coding, Mihaly Csikszentmihalyi’s concept of flow provides such a foundation. Through decades of research, Csikszentmihalyi studied artists, athletes, and scientists who became completely absorbed in their activities. These individuals described states where self-consciousness dissolved, time perception altered, and the activity itself became intrinsically rewarding.
Flow represents more than momentary satisfaction. Csikszentmihalyi distinguishes between pleasure, which requires no skill, and enjoyment, which emerges from activities that stretch capabilities. Watching television might be pleasurable but rarely produces flow. Solving a complex problem or playing challenging music can generate the deep satisfaction of optimal experience. This distinction proves crucial for education. If learning requires effort and growth comes through challenge, flow offers a model for making that effort intrinsically rewarding.
Three conditions facilitate flow. Clear, proximal goals provide essential structure. The individual must know what they’re trying to achieve moment by moment. “Become a better writer” is too abstract; “revise this paragraph to eliminate passive voice” provides necessary specificity. Immediate and unambiguous feedback creates tight loops between action and result. The basketball player knows instantly whether the shot went in. This feedback doesn’t always mean success; failure provides equally valuable information. Finally, the balance between perceived challenges and perceived skills makes up flow’s central condition. When challenges significantly exceed skills, anxiety results. When skills surpass challenges, boredom emerges. Flow exists where challenges stretch skills without overwhelming them.
The relationship between flow and learning extends beyond correlation. When students experience flow during learning activities, a fundamental shift occurs in their relationship to knowledge acquisition. Research shows that students reporting frequent flow experiences show superior performance across multiple metrics. They earn higher grades, demonstrate better retention, and exhibit greater persistence. These outcomes reflect the transformation of learning from an external obligation to an internal drive.
Flow-Teaching: A New Framework for AI Integration
Translating flow theory to AI-enhanced pedagogy yields something I would call “flow-teaching”—the deliberate orchestration of human-AI interaction to facilitate optimal learning experiences. In this model, AI serves not as an automated content generator but as a responsive collaborator that helps maintain the balance between challenge and support. The educator remains the architect of learning experiences, using AI to provide scaffolding while maintaining clear trajectories.
This framework acknowledges that successful application of AI, such as in coding or teaching, requires a focused approach to achieve particular objectives. A developer using AI to solve a technical problem engages in purposeful exploration, testing solutions against requirements and refining prompts based on results. Similarly, a teacher using AI effectively starts with clear pedagogical objectives and systematically explores how AI can support those goals. Both processes require expertise to evaluate outputs and iterate meaningfully.
Four operational principles form the basis of this framework. First, goal-directed design takes precedence over pure chance. Flow-teaching begins with intentional instructional design, not exploratory browsing. Educators identify specific learning objectives and then determine how AI might help achieve them. A mathematics teacher wanting students to understand derivatives doesn’t simply prompt AI for calculus problems. Instead, they identify conceptual milestones: understanding rate of change, connecting geometric and algebraic representations, recognizing patterns. After that, they create AI interactions to facilitate each stage.
Second, AI functions as a challenge-skill calibrator. The primary role of AI in flow-teaching is maintaining optimal challenge levels. This requires sophisticated application beyond adjusting problem difficulty. An AI supporting essay writing doesn’t simply correct grammar. It identifies the student’s current level and provides appropriately challenging feedback. For novice writers, it might focus on sentence structure. For advanced writers, it might challenge argumentative sophistication. The AI acts like a responsive sparring partner, adjusting intensity to maintain productive engagement.
Third, AI provides immediate and actionable feedback. Flow requires tight feedback loops, and AI’s capacity for instant response makes it valuable for maintaining flow states. But feedback must be both immediate and actionable. When a student submits code, effective AI doesn’t simply indicate bugs or errors. It explains why errors occurred, suggests debugging strategies, and provides hints without eliminating the challenge. This maintains forward momentum while preserving cognitive engagement.
Fourth, educators orchestrate the cognitive system and therefore retain ultimate responsibility for the learning environment. They select appropriate AI tools, design interaction protocols, monitor dynamics, and intervene when necessary. This includes assessing AI’s precision and its suitability for teaching, identifying when AI support helps or hinders learning, and finding a balance between AI use and human interaction. When AI suggests efficient solutions that bypass important struggles, teachers preserve productive difficulties. When feedback becomes formulaic, teachers provide encouragement. When students become overly dependent, teachers design activities reinforcing independence.
Practical Implementation
The distinction between vibe-teaching and flow-teaching has immediate classroom implications. In vibe-teaching, a teacher might prompt AI for a complete lesson plan, review it briefly, and deliver it largely unchanged. In flow-teaching, the teacher identifies specific conceptual challenges students face, designs activities addressing those challenges, and uses AI to generate varied examples maintaining appropriate difficulty. Surface activities might appear similar, but underlying logic differs fundamentally.
Student experiences also diverge. Under vibe-teaching, students use AI to complete assignments with minimal effort. Under flow-teaching, students engage through structured protocols, maintaining cognitive engagement. They might use AI to explore multiple solution paths, evaluate approaches, and justify choices. AI amplifies thinking rather than replacing it.
Assessment practices reveal another distinction. Vibe-teaching struggles with academic integrity as students submit AI-generated work. Flow-teaching reframes assessment around process documentation, critical evaluation, and creative application. Students might be assessed on their ability to improve AI outputs, identify limitations, or integrate assistance into larger projects. These assessments value cognitive skills developed through AI interaction rather than final products alone.
Consider a concrete example: a history class studying the Industrial Revolution. In a flow-teaching approach, the teacher designs an investigation where students use AI to analyze primary sources, generate historical perspectives from different social positions, and visualize economic data. The AI handles pattern recognition and generates viewpoints. Students provide critical analysis and evaluate competing narratives. The teacher orchestrates this interaction, ensuring accuracy, promoting critical thinking, and facilitating discussions connecting past to present. No single component could achieve these outcomes alone; power emerges from coordinated interaction.
Implications for Practice and Policy
The choice between vibe-teaching and flow-teaching reflects fundamental decisions about education’s purpose in an AI-augmented future. The vibe metaphor aligns with a transactional view where the goal is credential acquisition with minimal effort. Flow-teaching maintains education’s transformative mission of developing capabilities through meaningful challenge.
For practitioners, the flow-teaching framework offers immediate guidance. Rather than asking whether an AI tool saves time, educators should ask whether it maintains appropriate challenge, provides actionable feedback, and serves explicit goals. Teachers might begin lessons by stating objectives and explaining how AI interactions support them. They might design protocols requiring students to document interactions and evaluate suggestions. Assessment might focus on students’ ability to improve outputs and integrate AI-assistance into larger projects.
It is important to acknowledge the significant consequences for teacher education programs. Future educators need grounding in the learning sciences to understand flow conditions. They need technical literacy to evaluate AI capabilities. And they need design skills to orchestrate human and machine contributions effectively. This expanded preparation goes beyond prompt engineering to encompass the full complexity of creating optimal learning experiences.
Professional development for current educators requires similar attention. Most teachers today entered the profession before AI became available. They need support in developing orchestration skills through collaborative planning, peer observation focused on flow conditions, and action research examining how different applications affect engagement.
Clear research priorities also become apparent. We need studies examining how different interaction protocols affect flow states. Which patterns consistently maintain a proper challenge-skill balance? How do feedback timing and specificity influence engagement? Longitudinal research should investigate whether students learning through flow-oriented interaction develop different capabilities than those using AI primarily for content generation.
And finally, equity concerns demand particular attention. Flow-teaching requires sophisticated orchestration that may be unequally distributed. Well-resourced schools might implement it effectively while under-resourced schools could default to vibing. Further research should examine how to make flow-teaching accessible across contexts and identify minimal implementations preserving core principles while accommodating constraints.
Conclusion
As generative AI becomes increasingly powerful, the temptation to vibe through teaching and learning will intensify. The technology’s ability to produce adequate content with minimal effort creates an illusion of progress. But this path leads toward hollow education, maintaining formal structures while abandoning substantive development.
The alternative requires more effort but promises greater rewards. By replacing vague notions of vibes with the rigorous concept of flow, educators can harness AI capabilities while preserving education’s essential dimensions. Students engage in structured interactions, maintaining a cognitive challenge. And teachers orchestrate environments where technology serves pedagogical purposes. The result is an education both more effective and more satisfying.
Flow-teaching offers vocabulary for discussing AI integration with precision rather than metaphorical vagueness. It establishes criteria for evaluating tools based on pedagogical merit. Crucially, it prioritizes human expertise, judgment, and relationships within education.
Whether in software development or education, effective AI use involves directed exploration guided by clear goals and expertise. The conversation about AI in education has only begun. As these technologies develop, we need conceptual frameworks to guide responsible integration. Flow-teaching represents one such framework, grounded in research and aligned with principles. By adopting it, the educational community can move beyond vibing’s superficial promise toward approaches truly serving learning and development. The choice we make now will shape not only how we teach but what we value in human development for generations to come.
How are you navigating AI in your teaching or learning practice? Have you caught yourself “vibing” through lesson planning when pressed for time? What protocols have you developed to keep students thinking rather than just prompting? For students: how do you use AI while ensuring you’re actually learning? And for those who’ve found effective frameworks beyond trial and error—what’s working? Share your experiences and strategies in the comments.
P.S. I believe transparency builds the trust that AI detection systems fail to enforce. That’s why I’ve published an ethics and AI disclosure statement, which outlines how I integrate AI tools into my intellectual work.
The terms ‘vibe-teaching’ and ‘vibe-learning’ as used in this essay should not be confused with educational organizations or services that use ‘vibe’ as part of their branding.