The Evolution of AI Value

The first wave of generative AI focused primarily on content creation - ChatGPT writing articles, Midjourney generating images, essentially replacing traditional production roles. However, as these technologies mature, their greatest value might well shift towards distribution and personalization rather than raw production.

From RSS to Recommender Systems

The evolution of content distribution reveals how technology repeatedly transforms information access. RSS (Really Simple Syndication) represented an early attempt to solve content discovery, providing a pull-based system where users subscribed to feeds they cared about.

Personal Note: I interned at China Impact Investing Network (CINN), down the road from Huangzhuang, earning 100 RMB daily for translation and RSS-related tasks - work that would now be largely automated by GPT.

As content volume exploded, the focus shifted to algorithmic distribution through recommender systems. These attempted to match existing content to user preferences through increasingly sophisticated methods, but still fundamentally operated on a “create once, distribute many times” model.

Content Creation vs. Distribution Costs

The economics of content have always been defined by the balance between creation and distribution costs, as illustrated in our visualization. Traditional models rely on two fundamentally different approaches:

Professional vs. User-Generated Content Economics

Traditional PGC platforms invest heavily in upfront content creation ($200-500 per article) while optimizing distribution costs ($0.001 per user). This creates an economic model where high-quality, centrally produced content must reach massive scale to be sustainable. Users remain passive consumers with minimal influence on content direction.

UGC platforms invert this model by outsourcing creation costs to users while investing primarily in discovery infrastructure. This creates greater diversity but inconsistent quality. Both approaches increasingly allocate resources to distribution rather than creation as competition for attention intensifies.

The discovery paradox emerges: as content volume increases, the marginal value of new content approaches zero without effective discovery mechanisms. Users face decision fatigue from too many choices, and the market naturally shifts investment toward discovery rather than production.

Generation as Distribution: Personalized Content at Scale

AI fundamentally changes this paradigm by collapsing the distinction between production and distribution. When content can be generated at the moment of consumption, personalized for each user, the model shifts from:

Traditional: Create once → Distribute to millions
Generative: Create parameters → Generate millions of variations

The dual-axis economics chart reveals this transformation. Traditional content scales efficiently after high initial investment but delivers mediocre value. Generative approaches provide dramatically higher user value but face linearly increasing costs that become prohibitive at scale.

The optimal approach emerges through content modularity - recognizing that not everything needs regeneration. By identifying which components provide the most personalization value, hybrid approaches can maintain 80% of personalization benefits at 30% of the cost.

The 60/20/20 Rule: Strategic Content Modularity

Most knowledge domains contain substantial core material that remains consistent across users. The 60/20/20 rule maximizes efficiency by segmenting content into:

  • 60% static core content (foundational principles, established facts)
  • 20% cohort-level content (industry examples, cultural contexts)
  • 20% individual personalization (connections to personal experience, learning pace)

This approach creates a fundamentally different economic curve that scales more efficiently than traditional content while maintaining most personalization benefits. For a business book distributed to 50,000 professionals, this approach can deliver twice the relevance at the same cost as traditional publishing.

Multi-modal Personalization: Beyond Text

The personalization paradigm extends beyond text to encompass multiple modalities. Content can dynamically transform between formats - text summaries becoming virtual presenter videos, complex topics converting to interactive explanations, news transforming into personalized audio briefings. This multi-modal capability increases generation costs but dramatically enhances engagement and information retention.

These cross-modal transformations add approximately 30-40% to generation costs but can increase engagement by 200-300%, creating compelling economics despite the higher production expense. Each user’s preferred learning style becomes another dimension for personalization.

The User Advocate: Beyond Algorithmic Recommendation

The most powerful personalization emerges not from content formatting but from deeper user understanding. The User Advocate concept represents an AI persona that truly comprehends the user’s interests, knowledge level, and perspective, then guides content creation accordingly.

Unlike recommendation systems that rely on sparse signals, the Advocate builds a comprehensive user model through conversation and observation. This enables exploration of “unknown unknowns” - valuable topics users didn’t know to search for. The approach fundamentally changes platform economics by aligning incentives with actual user satisfaction rather than engagement metrics.

Fluid Knowledge and the Future

The most significant impact of AI lies not in replacing content creators but in transforming how knowledge flows to individuals. As the internet solved information scarcity, generative AI now solves the problem of relevance through “fluid knowledge” (知识液化) that adapts perfectly to each person’s context.

In this emerging paradigm, content becomes transformable across formats, users experience the feeling of being truly understood, and exploration replaces search as the primary discovery model. The User Advocate becomes a critical interface between vast information spaces and human understanding, fundamentally changing our relationship with knowledge acquisition.

Product Prototyping

You notice a card about recent advancements in your industry. As you engage with this content, your User Advocate (an AI persona named Alex that you’ve configured to match your learning style) notices your particular interest in one aspect and suggests a tangential topic you hadn’t considered - an “unknown unknown” that connects surprisingly well with a project you mentioned in a previous conversation.

When you switch devices during your commute, the same content transforms seamlessly into an audio format, picking up exactly where you left off but adapting to the new context. Later, when discussing the topic with colleagues, KnowledgeStream generates a shareable summary that maintains the personalized framing that made it valuable to you, while adapting select elements to resonate with your team’s shared context.

The system’s economics work because it’s not generating everything from scratch. The core industry data, fundamental concepts, and base analysis remain consistent, while personalization happens strategically where it adds the most value. The result is information that feels bespoke without requiring the computational resources of complete regeneration.