Quantum 1.0

2026-09-25

Quantum 1.0 is out. This post says what it is, what the version number promises, how we checked that the promise holds, and what comes next.

What Quantum is

Declarative web applications in XML, with AI and RAG built into the language. No build chain, no JavaScript, no front-end framework.

A page is a .q file. State, parameterised SQL, forms and their validation, components, authentication, file uploads and mail are tags; so are an LLM call that answers from a knowledge base and cites its sources, and an agent whose tools are written in Quantum itself. quantum start serves a folder of pages; there is nothing to compile.

It is for solo developers and small teams building internal tools, dashboards, admin panels and AI applications who would rather not maintain a front-end build chain. It is not a general-purpose replacement for a JavaScript framework.

What 1.0 promises

The promise is written down in Support tiers, and the engine enforces it:

• Core — components, q:set, q:if, q:loop, q:function, q:query and q:transaction, q:action and forms, q:invoke, q:data, composition, files, mail, the Core set of ui:*, authentication.
• AI — q:llm, q:knowledge, q:agent.
• Experimental — tags that run but carry no promise (jobs, messaging, websockets, Python scripting, the terminal target, …). A tag outside Core and AI prints a warning the first time it runs.
• Laboratory — the 2D game engine and other targets kept to exercise the language, with no promise at all.

From 1.0, Core and AI follow semantic versioning: a 1.x release does not break a program that uses only them. A change that would waits for 2.0. Experimental and Laboratory may change in any release.

How it was checked

A promise is only as good as what checks it. Four things do:

1. A specification with rule IDs. SPEC.md states what a Quantum program means in 137 numbered rules (LOOP-2, DB-4, IA-6…), and it is frozen at 1.0. Every rule is cited by at least one conformance test — 866 of them — and CI fails if a rule has no test or a test cites a rule that does not exist.
2. Apps that use it. The repository has small, real apps for each tier: a task list drawn with ui:* in the browser, the terminal and a desktop window; a blog; a helpdesk with attachments and mail; a bank transfer inside one q:transaction; a documentation assistant and a shop agent for the AI tags.
3. quantum test. Apps are tested in their own language — a *.test.q visits a page, submits an action and checks the redirect, the flash, the table rows and the history, each test on a fresh database built from the migrations. The apps' suites run in CI.
4. Live AI tests. The AI tags are tested against a real model server before every release, with structural assertions (the tool was called, the right chunk was retrieved), never exact text.

And the status page is not written by hand: a script runs every example and reports what parses and what executes.

What changed on the way

The road from 0.9 to 1.0 was mostly about making the promise true: behaviour that used to fail silently — a misspelt name, an attribute that did nothing, a missing key, a query field that does not exist — is now an error that points to the line. The changelog has the details, version by version.

What comes next

• quantum test grows: more of what an app does should be checkable in its own language.
• Experimental → Core, one tag at a time, and only with a SPEC rule, a test that cites it, an example that runs and a guide page.
• This site: translations and a domain of its own.

Bugs, questions and ideas are welcome on GitHub Issues. If Quantum is useful to you, the support page says what help pays for.