Every framework war comes to an end. But what if the winner of that war is the one who determines how new code is created going forward?
I can't stop pondering why React won't die. Not the "it's still sufficient" kind of way, but in the "it might be structurally impossible for it to lose" kind of way.
The Training Data Moat
Because LLMs are trained on React code at such a higher rate than other frameworks. It's not a conspiracy, it's just math. React has been the dominant framework for nearly a decade. That's a decade of stack overflow answers, blog posts, open-source repos, and tutorials being fed into the machine.
When a junior dev asks GPT to scaffold a component, they get React. When a senior dev uses copilot to autocomplete a hook, the suggestion is based on React patterns. The output reinforces the input. 🔄
New Developers Never Get the Choice
This is what I find concerning. Many new developers these days use AI-generated boilerplate which is based on React patterns. They haven't actually made a conscious choice to use React after considering other options. React is chosen for them by AI, unless they specifically ask for alternatives.
→ Ask an LLM for a todo app — you get React
→ Ask for a dashboard layout — you get React
→ Ask for "a modern web app" — you get React
And that's not a fair competition.
The Feedback Loop Nobody Talks About
What this results in is that with more React code in the training data, you get more React recommendations. With more React recommendations, more React code is generated. With more React code, you get even more React in the next training data.
It's a loop that doesn't really consider the technical aspects of the framework.
It's great that there is TanStack Start, Svelte, and Solid. They are all really good options. Some might argue that they are even better for certain scenarios. But React is often the default choice generated by the AI. And their market share compared to React is minuscule – and potentially shrinking because of this vicious cycle.
Merit Stopped Mattering
I'm not saying React is not good – it is. But these other frameworks are good too, and for some specific scenarios, they might even be better. But “good” should not be an adequate reason for one framework to dominate the scene.
It’s rather frightening in that, authors of a new framework could objectively ship something better and still lose. Not because developers tried both and chose React. The AI middleman just never presented the alternative. 😬
It’s a new kind of lock-in. Not vendor lock-in. Not ecosystem lock-in. Cognition lock-in. The tool we think about code with has already started the game with one-nail filled bat buried in the backyard.
What Breaks the Loop?
What breaks this loop?
Honestly? In the short term, I’m not sure anything does. Perhaps one of the big LLM players will deliberately diversify their training data. Perhaps asking “which framework” becomes normal for framework agnostic AI. Perhaps someone will build an AI coding tool that asks “which framework?” before deciding. None of those things are happening right now. Right now, every AI-assisted line of code is a vote for the status quo.
React will not die, it has achieved immortality. It will also not die because the robots writing our code were brought up on it. This is not a meritocracy, this is a feedback loop that has a meritocracy’s t-shirt on. 🤷
So here's my question: if you're building something new today and you deliberately pick a non-React framework, how much harder does AI tooling make your life? Has anyone felt this friction firsthand?
Top comments (0)