Proofed, pressed,
and ahead.

The .pyxl format: async loader, server actions, and the React component in a single file — server-rendered, hydrated, hot-reloading, with file-based routing.

Declare plugins in pyxle.config.json and they wire services, middleware, and their own schema at startup. WebSocket endpoints and the client navigation cache landed alongside.

Edge caching with per-route TTLs, sanitized production errors, layout loaders, a faster static path, and live route-shape changes with no restarts.

pyxle serve --workers N runs an independent server per core with small assets served from memory. HEAD output moved behind a sanitising allowlist and CSRF exemptions began matching on path boundaries.

The wave that took Pyxle from rendering pages to running apps: Pydantic-validated actions with OpenAPI output, streaming SSR through <Suspense>, caching with SSG and ISR, WebSocket pub/sub, background work, built-in observability, a Next.js-parity <Image>, a rate limiter, and a type-aware JSX side.

Enable the llms feature and every page answers at its URL with .md appended, or to an Accept: text/markdown request — plus a generated /llms.txt index and discovery headers. This page has a twin; so does every doc.

A release forged by testing: React 19 and Vite 7 in new projects with pyxle init asking what to include, concurrent streaming SSR with per-request isolation, a calmer dev server, shadcn/ui wired end to end.

A dev-only dashboard at /__pyxle/studio built into pyxle dev — routes, an interactive tester, a live request feed, metrics, resolved config — and breakpoints that bind on both sides of a .pyxl file at once, with tracebacks that finally name your source.

pyxle-db (SQLite, PostgreSQL and MySQL behind one explicit-SQL API), pyxle-auth (argon2id sessions, RBAC, API tokens) and pyxle-mail (SMTP, Resend or console) — every plugin tested in CI, the database claims against live PostgreSQL and MySQL.

Reproducible two-box benchmarks against six frameworks, the load generator on its own hardware, and the losses shown next to the wins.

A launch program: the first five community plugins that meet the published standards are recognised as ecosystem pioneers, with the spotlight and a roadmap seat. The bar is public; the race is to the bar, not to the submit button.

The Phase B design discussion is open — how plugins will ship their own API endpoints, then whole pages. Founding authors co-design it.

pyxle-auth providing its own sign-in endpoints out of the box; webhook plugins owning their URLs with the right CSRF posture.

Sign-in screens, dashboards — and the admin-panel moment this framework should eventually have.

The DatabaseLike idea, generalised: framework-defined interfaces for mail, storage, cache, and more, so providers (Postmark, SES, S3, R2…) are swappable by config, not code. Pyxle ships the interface; the community ships the adapters — one recommended pick per capability, alternatives behind it.

Mail, storage, jobs, observability — the plugins the ideas page calls for, built by the community or by us.

The type-aware JSX side — generated .d.ts declarations and a pyxle typecheck gate — shipped in v0.5. Next is writing .tsx directly in your .pyxl, with loader-data types generated from the Python side: the request that comes up most, answered end to end.

The overlay the debugger did not ship: click any element on screen and jump to the .pyxl that rendered it, plus a bundle-size treemap for the client build.

Diagnostics, completions, and go-to-definition that cross the Python↔JSX boundary; pyxle check growing into a real linter.

Engineering notes and honest write-ups — how the SSR pool works, what the benchmarks taught us, what broke.

The release where minor versions stop being allowed to break you. It lands when the plugin API has proven itself, not on a date.

Managed hosting from the people who write the framework: push a Pyxle app and skip the server story entirely — with AI scaffolding in the mix. In development; the newsletter hears it first.