Solid 2.0 reaches stable with async in the reactive graph

Solid 2.0 reaches stable with async in the reactive graph

The Solid project has announced that Solid 2.0 has reached stable. The post, titled "The Long Hold", is written in the first person by an author who is not named in the text. He says it took 5 years from his first commit of Solid to 1.0, and another 5 from 1.0 to now. Solid 1.0 bet on fine-grained reactivity (signals) as the foundation for user interfaces. The post says that position was once unusual, and that signals now appear in Angular, Svelte's runes, Vue Vapor and a proposal before TC39 to put them in the language itself. The author also says the conversation reached React's compiler.

The first headline feature is async inside the reactive graph. A computation can return a Promise or an async iterator, and everything downstream handles it. The author says this automates race-condition handling, that introducing async does not force an app refactor, and that speculative state is rolled back automatically so the server stays the authority on truth. Server functions, which SolidStart introduced in early 2022, now live in Solid's core. They support RPC calls over Request/Response and persistent connections over Server Sent Events. The author says new features, such as reactive Server Components, will be announced in the coming weeks; no details are given.

The second theme is writing code with agents. The author anticipates the objection "But, we aren't writing the code anymore!" and argues that a system built on constraints produces less code to maintain and fewer edge cases. He says most frameworks claim to "work great with AI", which usually means only that models have seen a lot of their code, and that the real question is what happens when the agent gets it wrong. In most frameworks a reactive mistake is silent. Solid's graph knows what was read, where, and what will update, so its development builds report mistakes as they happen, with a stable code, the location in the source and the fix. Because async is in the graph, Solid can also detect waterfalls, poorly constructed effect loops and poorly scaling graph shapes such as wide fanouts and fan-ins. The team says it has integrated with new dev tooling and Chrome's Performance Tracks, and the same observability powers an updated Sentry plugin so production builds can report the same information. The area is still evolving as the team learns how agents build with Solid. Solid 2.0 also adds experimental support for ReactiveScript and TSRX, both aiming to use compilation to improve agent authoring further.

The third theme is ecosystem structure. The author says React's strength was composability, but that its missing primitives fractured its ecosystem. Solid wants to give developers primitives to build on, such as server functions and support for any router and cache solution for Single-Flight Mutation. On that basis the team retired its metaframework SolidStart, saying every primitive needed to build a Solid metaframework is now provided by core or by framework-agnostic Vite plugins. Core supplies Request/Response primitives built on web standards and hooks for observability, so Solid 2.0 needs only one Clerk or Sentry plugin. The author says TanStack Query and Router have already proved as capable as anything built first party, and he credits Tanner's attention to detail on APIs.

The fourth theme is a smaller API. Despite more capability, the post says the surface area has collapsed almost everywhere. createResource, batch, startTransition, on, produce, createComputed and createMutable are all gone. The author says equivalent code is shorter and clearer on intent, and that fewer escape hatches such as createEffect make it "safe by construction rather than by exhaustion".

On the road to stable, the author says discussion of 2.0 began the moment 1.0 shipped, the real effort started in late 2024, and the road to 2.0 was announced in February 2025. A year of experiments followed, including rewriting the reactive core from scratch until the async model held up. The beta landed in March and the Release Candidate in August, with the API frozen. Then came 23 release candidates in 14 weeks, with 971 pull requests merged and 753 issues closed. The bar for stable was a full week with no regression in ordinary app code and no crash or hang anywhere. By the end, the author says, bug reports came from people (agents) deliberately hunting edge cases in the deepest corners of the scheduler. He thanks contributors and sponsors, including Sentry, Netlify, Google Chrome, JetBrains, Anomaly, Cursor, Code Rabbit, Anthropic and Tembo, and says the past two and a half years have been some of the most difficult of his life. The title refers to a diagnostic warning in 2.0, the Long Hold, which fires when a hold has been shown (isPending) but enough time has passed since the last input.

Key facts

  • Solid 2.0 is stable: computations can return a Promise or async iterator, and downstream code handles it, with speculative state rolled back automatically.
  • Server functions, introduced by SolidStart in early 2022, now live in core and support RPC over Request/Response and persistent connections over Server Sent Events.
  • Development builds report reactive mistakes with a stable code, the source location and the fix; the aim is to guide AI agents, and the same observability feeds an updated Sentry plugin.
  • SolidStart is retired, and seven APIs are removed: createResource, batch, startTransition, on, produce, createComputed and createMutable.
  • The release candidate phase ran 23 RCs in 14 weeks, with 971 pull requests merged and 753 issues closed; the stable bar was one week with no regression and no crash or hang.

Why it matters

Solid helped popularise fine-grained reactivity, and the post says signals now appear in Angular, Svelte's runes, Vue Vapor and a TC39 proposal. A stable 2.0 is the project's answer to what comes next: async handled in the reactive graph rather than around it, and a much smaller API. It also stakes out a position on AI coding. The author argues that a framework which knows what was read and what will update can flag agent mistakes that other frameworks leave silent, which he contrasts with claims that a framework "works great with AI" merely because models have seen a lot of it.

Who it affects

Developers using Solid 1.x face a real API change, since createResource, batch, startTransition, on, produce, createComputed and createMutable are gone. SolidStart users are affected by its retirement. The source says only that it was retired and does not describe what happens to existing SolidStart apps. Library authors for auth and monitoring, such as Clerk and Sentry, are told they now need to ship one plugin. Teams letting agents write their UI code are the audience for the new diagnostics.

How to use it

The post itself is an announcement, not a tutorial, and its code samples are not included in the text. What it does say: author async by returning a Promise or async iterator from a computation; use server functions from core; in development builds, read the reported diagnostics, each with a stable code, a source location and a suggested fix; and use the updated Sentry plugin to get the same information in production. Experimental ReactiveScript and TSRX support is available but not stable. No migration guide or upgrade path from 1.x is described.

How solid is it

This is a first-party announcement from the Solid project, written by an author who is not named in the text, so the claims are the project's own. The process figures are concrete: 23 release candidates in 14 weeks, 971 pull requests merged, 753 issues closed, and a stated bar of one week without regression or crash. No benchmark, bundle-size or performance figures are given. No evidence is given that the diagnostics improve agent output; that is the author's claim only. The sponsors listed are thanked for supporting development over the years, and they are not stated to have endorsed this release.

Risks and caveats

No exact release date or version number is given. The diagnostics area is described as still evolving, and ReactiveScript and TSRX are experimental. Reactive Server Components are only promised for the coming weeks, with no detail. Removing seven APIs and retiring SolidStart means existing apps will need changes, and the post does not say how. The author also describes the work as long and personally costly, and acknowledges in the closing note on the Long Hold that it took a while for the async work to land.

“Every framework now says it "works great with AI", and usually that means models have seen a lot of it. That's a statement about training data, not about the framework.”

— Solid 2.0 announcement post, The Long Hold