What version of Hono are you using?

4.6.8

What runtime/platform is your app running on?

Bun

What steps can reproduce the bug?

type UserContext = {
  Variables: {
    userId: string
  }
}

// Works as expected - both middlewares maintain correct type
const app1 = new Hono().get(
  "/ws",
  createMiddleware<UserContext>(async (c, next) => {
    c.set('userId', "12345")
    await next()
  }),
  async (c, next) => { // c is correctly inferred as Context<UserContext, "/ws", {}>
    await next();
  }
);

// Type inference breaks in the middle middleware
const app2 = new Hono().get(
  "/ws",
  createMiddleware<UserContext>(async (c, next) => {
    c.set('userId', "12345")
    await next()
  }),
  async (c, next) => { // c is incorrectly inferred as Context<BlankEnv, "/ws", {}>
    await next();
  },
  async (c, next) => { // c is correctly inferred as Context<UserContext, "/ws", {}>
    await next();
  }
);

What is the expected behavior?

The UserContext type should be maintained throughout the entire middleware chain, regardless of the number of middlewares.

What do you see instead?

In a chain with more than two middlewares, intermediate middlewares lose the custom context type and revert to BlankEnv, while the last middleware maintains the correct type.

Additional information

It's possible to type the Hono instance itself with new Hono<UserContext>().
While this approach fixes the type inference, it introduces a type safety issue. The variables become accessible in all middlewares, even before they are actually set. For example, if userId is only set in the second middleware, the type system would incorrectly allow access to it in the first middleware where it hasn't been initialized yet.