React Compiler: Rust edition is coming soon. We've ported the majority of the passes using AI. When the initial port finishes we'll do some updates to get the code in a state we're happy to maintain, then extensive testing and look at performance. More to come soon
There are so many takes on the "give the agent a DAG of tasks" idea. But the reality is that task lists were always an attempt to impose order against fundamentally chaotic systems. You find the dependencies by doing tasks as often as you realize them in advance.
A thing that you can do, when seeing a technical choice you disagree with: pause. Take a breath. Be curious. Ask yourself, "why might they have done that?". Ask the authors why they did that? What context do they have that you don't?
I'm really curious for feedback about React Compiler's error messages. Do people prefer very short ESLint-style messages ("can't do X") w no context/information, or informational messages with more information?
Example: playground.react.dev/#N4Igzg…
More context in the discussion post at github.com/eslint/eslint/dis…
We've been pretty happy with TypeScript for React Compiler, but actual pattern matching (with exhaustiveness checking by default!) is so good it makes me want to port to Flow: flow.org/en/docs/match/
For compiler nerds: a writeup of the core inference rules and approach that React Compiler uses to decide which groups of instructions to memoize together. If something isn't clear let me know! github.com/facebook/react/bl…
Joe Savona retweeted
We're thrilled to announce that @sethwebster @shrutikapoor08 @en_JS @zmofei @lydiahallie @jherr @lunaleaps will be speaking at React Conf 2025!
The thing I loved about Create React App was that you could quickly install, hack on a single component, and quickly see changes. It was a great way to learn React. *All of the frameworks we recommend provide that exact experience*. Install, hack on the root component, iterate.
Call me old fashioned, but I believe that people convicted of violently attacking police officers belong in jail.
19rc1 is out with a new feature, sibling pre-rendering, that is the best of both behaviors. Amazing work by @acdlite and @sebmarkbage to implement this feature and @rickyfm, Jack Pope, and the rest of the team for verifying and rolling it out github.com/facebook/react/is…
This is harder than it seems (always is) because of the size of our community & many ways people use React. We need to balance info overload while getting a diverse, representative set of feedback.
We also don't want to privilege established players at the expense of making it hard for new libraries/frameworks/educators/voices to emerge. Rushing into a fix could backfire and lead us down the wrong path.
The result was that we didn't get the feedback until the 19 RC. Our changelog did call this out (could have called it out more prominently, to be fair), and that's how folks knew to test the behavior. Once we heard the feedback we quickly discussed, leading to where we are.
When using this pattern, it was inefficient to try to render other siblings of a suspended child, since we would have to show the fallback in either case. So for React 19 we changed to show the fallback right away, then render again once a child un-suspends.




