Dead Framework Theory (aifoc.us)

"Dead Framework Theory" argues that React has effectively become the web platform, not merely one framework among many, because of a self-reinforcing feedback loop between LLM training data, developer tooling, and site output. Observations: developer-facing tools (Replit, Bolt, etc.) often hardcode React in system prompts; LLMs increasingly emit React by default; millions of new sites built via LLM-assisted tooling are React-based; and token-usage curves (e.g., OpenRouter’s billions/day) line up with React’s growth. Dataset checks (HTTP Archive vs BuiltWith, top-1M comparisons) are noisy but consistent enough to suggest React’s adoption keeps accelerating while competitors stagnate. For the AI/ML and web ecosystems this matters because LLMs now shape what code is discoverable and maintainable: if a framework or new web API isn’t present in training corpora or tool prompts, it won’t be output (estimated 12–18 month lag to surface). That creates high barriers for new frameworks, libraries, and platform features unless they (a) get into LLM training data, (b) convince toolmakers to change system prompts, and (c) build a sprawling library ecosystem — all while competing with entrenched React patterns. Practical implications: expect slower adoption of new APIs (CSS Nesting, platform features) unless they provide capabilities impossible in user-space (WebGPU, Passkeys, multi-page transitions), and consider new go-to-market or paid-integration models to seed tooling and training corpora.