AI Is Crushing the Knowledge Economy: The Truth Behind Course Sales Dropping by Half

Article

Prominent front-end educator Josh W. Comeau recently shared an observation that sparked widespread discussion across the industry: his third course, Whimsical Animations, sold only one-third as many copies as his previous courses typically did at launch. His existing two courses have also seen significant year-over-year revenue declines. And he's not alone — after speaking with multiple course creators, he found everyone experiencing the same trend: revenue down more than 50%, engagement dropping, and users migrating en masse to large language models (LLMs).

This brief post on Bluesky cut straight to the heart of a painful reality facing the technical education industry and the broader knowledge economy. It reflects the deep structural impact AI is having on the business models of content creators.

Background on the knowledge economy: The paid knowledge industry took off around 2016 in China before scaling into a global market. Platforms like Udemy, Coursera, and Teachable, alongside independent creators like Josh W. Comeau, form a multi-billion-dollar technical education ecosystem. Before the AI wave hit, top technical course creators could earn millions of dollars annually, with quality courses typically having a three-to-five-year lifespan. Comeau himself built a stellar reputation in the developer community through his accessible React and CSS courses, with pricing usually ranging from $100 to $300 — placing him among the top tier of independent creators. That context makes his declining numbers all the more significant: if even the best creators are feeling the impact this acutely, the rest of the industry is in an even more precarious position.

The AI "Double Whammy"

Comeau attributes the sales decline primarily to AI, describing it as a "double whammy."

Hit #1: Career Anxiety Suppresses the Will to Learn

The first blow comes from shrinking demand. As AI coding capabilities have advanced rapidly, many people are asking themselves: "Will developer jobs even exist in a few months?" This uncertainty about career prospects directly undermines the motivation to invest time and money learning new development skills.

When professionals in any field start doubting their future, "self-investment" spending is typically the first to get cut. Learning a new framework or mastering a new animation technique feels far less urgent when the overriding anxiety is "will I even keep my job?" It's a rational market response — but a brutal one for educators.

It's worth noting that this career anxiety isn't unfounded. AI coding tools like GitHub Copilot and Cursor have been shown to significantly boost developer productivity, with some research indicating efficiency gains of 30% to 55%. This means the same product can potentially be built with fewer developers — a trend already visible in the tech industry layoffs of 2023–2024, where many companies explicitly cited "AI replacement" as a reason for reducing headcount.

Hit #2: LLMs Become Free Personal Tutors

The second blow is more direct — even for those who still want to learn new skills, LLMs now provide personalized one-on-one instruction, dramatically reducing the need to purchase paid courses.

In the past, the value of a well-structured, progressively designed course lay in its systematic organization of knowledge and the distilled expertise of the instructor. Today, learners can ask ChatGPT or Claude anything at any moment, get instant, targeted answers, and even have AI tailor a learning path to their specific skill level. When an "on-demand AI tutor" is essentially free, paid pre-recorded courses priced in the hundreds of dollars face enormous pricing pressure.

LLMs as learning tools — technical background: The conversational learning capability of large language models stems from pre-training on massive text corpora and fine-tuning via RLHF (Reinforcement Learning from Human Feedback). Models like ChatGPT and Claude can dynamically adjust the depth of their explanations based on a user's specific question, code context, and skill level — an individualization that technically approaches the effectiveness of traditional one-on-one tutoring. Compared to pre-recorded courses, LLMs' core advantage is "zero-wait Socratic dialogue" — learners don't need to scrub through video to find answers; they describe their confusion directly and receive immediate feedback. According to OpenAI data, coding and technical queries represent a substantial share of total ChatGPT usage, confirming that developers are already treating it as their primary learning tool. The capabilities of next-generation models like GPT-4o and Claude 3.5 Sonnet for code comprehension and instructional explanation far exceed what early versions could do two years ago, making their substitution effect increasingly pronounced.

A Deeper Ethical Dilemma: Knowledge "Consumed" Without Consent

One line from Comeau's post stands out as particularly pointed: the LLMs people are turning to "slurp up all of our work and regurgitate it, without consent or compensation."

This cuts to the core contradiction in current AI ethics debates. Those excellent technical tutorials, blog posts, and open-source code repositories are precisely the kind of material used to train large models. Educators' painstakingly created content gets "learned" by AI and absorbed into model capabilities — and then that same AI turns around and takes their business.

AI training data and copyright disputes: The copyright controversy around AI training data has evolved into several major lawsuits, signaling that the issue has moved from moral debate into legal territory. The New York Times sued OpenAI and Microsoft in late 2023, alleging unauthorized use of millions of news articles for training; multiple visual artists filed class-action suits against Stability AI, Midjourney, and DeviantArt; and GitHub Copilot faces copyright challenges from the open-source developer community, with the central dispute being whether it directly reproduces copyright-protected training samples in generated code. The core legal question is whether training a model on copyrighted content constitutes "fair use" under U.S. copyright law. U.S. courts have yet to establish binding precedent, while the EU AI Act has taken an early step by requiring AI companies to disclose training data sources. This legal vacuum is one of the fundamental reasons creators like Comeau feel powerless.

This creates a deeply troubling closed loop:

• Creators produce high-quality content
• AI companies scrape that content for training
• The trained model displaces demand for the original content
• Creator revenue falls, reducing the incentive to create

This isn't just a business problem — it's a profound challenge to intellectual property, data consent, and the fair distribution of value. When freely available, openly shared high-quality content becomes fuel for AI, but creators receive nothing in return, is this model sustainable in the long run?

Industry-Wide Disruption in the Knowledge Economy

Comeau is emphatic that this isn't his problem alone. "I've spoken with a number of course creators and we're all seeing the same trends." This points to a structural industry issue, not a failure of any individual product or marketing strategy.

Structural changes in the technical education ecosystem: The contraction of the technical education market extends beyond course sales into the broader knowledge content ecosystem. Revenue from technical blog advertising, watch time on YouTube tech channels, and subscription rates for coding newsletters have all declined to varying degrees. The most telling data point comes from Stack Overflow — once the world's most important developer Q&A community — which reported a roughly 14% year-over-year traffic decline in 2023 and subsequently announced a 28% workforce reduction. The industry widely views this as strong evidence of a mass migration by developers from traditional information sources to LLMs. Meanwhile, Udemy and Coursera earnings reports show increasing pressure on paid conversion rates for individual technical courses. This structural shift means the disruption isn't confined to any single content format — the entire business model of "delivering technical knowledge through text and video" is facing a systemic revaluation.

This observation was quoted by tech blogger Salma Alam-Naylor in a post titled Goodbye Forever (Probably) — a title that captures the complex and resigned emotions content creators feel in the face of the AI wave. The entire technical education ecosystem seems to be standing at a critical inflection point.

Where Can Content Creators Pivot?

While Comeau's post primarily describes the current situation, we can extrapolate potential paths forward for knowledge economy creators:

• Lean into community and accompaniment: AI can provide answers, but it struggles to replicate genuine human community, peer feedback, and emotional connection.
• Offer experiences AI can't replace: Live workshops, hands-on project mentorship, personalized code reviews, and similar high-touch formats.
• Reposition the value of content: Shift from "delivering information" to "providing taste, judgment, and carefully curated learning paths" — areas where AI currently underperforms.
• Explore new monetization models: Subscriptions and paid communities rather than reliance on one-time course sales.

Closing: A Mirror for an Entire Industry

Comeau's post resonated so widely because it uses the most direct data and firsthand experience to document AI's real impact on a specific industry. Revenue cut in half, engagement collapsing, work absorbed without compensation — these aren't abstract predictions about the future. They're things happening right now.

For technical educators, content creators, and all professionals who rely on monetizing knowledge, this is a signal that must be taken seriously. AI is empowering learners while simultaneously reshaping the entire knowledge value chain. Finding a new foothold in this transformation — and advocating for fairer rules around AI data usage — will be among the defining challenges of the years ahead.