Introducing Zirn v1.4 — Raxat

Knowledge work has a strange problem.

We spend years collecting information—lectures, PDFs, articles, research papers, highlights, notes, ideas—and then organize them into folders that gradually become harder to navigate than the information itself.

The more we learn, the more fragmented our knowledge becomes.

Most software accepts this tradeoff.

We don't.

Today we're releasing Zirn 1.4, codenamed Raxat.

Raxat is built around a simple belief: your knowledge should become more useful as it grows, not less.

To move toward that future, we've rebuilt major parts of Zirn's foundation.

Users can now route different cognitive tasks to different AI systems. Content generation, flashcard creation, and Home generation are no longer tied to a single model. Instead, they can be independently configured based on the strengths of each provider.

That matters because knowledge work is not one task. Reading a source, extracting structure, drafting an explanation, generating flashcards, and building a workspace Home view all ask for different kinds of reasoning. Raxat gives those jobs room to be tuned separately, so Zirn can become more capable without forcing every workflow through the same narrow path.

We've also expanded Zirn's model ecosystem with DeepSeek support, while continuing to use dedicated document-reading infrastructure for OCR and source extraction.

But Raxat isn't primarily a model release.

It's a workflow release.

Writing remains one of the most important ways humans think. That's why Raxat introduces a redesigned editing experience, native Undo and Redo support, persistent highlighting, and a cleaner interface that keeps attention on ideas instead of controls.

The editor now feels more like a place to stay. Highlights survive the flow of reading and revision. Undo and Redo behave the way a native Mac app should. The surrounding interface has been pared back so the important work—the sentence, the source, the connection you're trying to preserve—stays in front.

Learning received similar attention.

Flashcards are no longer temporary outputs generated and forgotten. They now live inside a dedicated workspace and are cached intelligently using source fingerprints and similarity checks. In many cases, Zirn can reuse existing learning material instead of spending additional tokens regenerating knowledge it already understands.

This changes the role of flashcards inside Zirn. They are not disposable side effects of an AI request. They are part of the knowledge base: reviewable, reusable, and connected to the sources that produced them. As the workspace grows, learning material can grow with it instead of being recreated from scratch.

“Knowledge should not merely accumulate. It should become easier to return to, easier to trust, and easier to grow.”

The result is a system that feels less like a collection of AI features and more like a coherent knowledge environment.

Throughout development we found ourselves returning to the same question:

What happens when software stops treating information as documents and starts treating it as knowledge?

Raxat is another step toward answering that question.

Today, Zirn can help organize information.

Tomorrow, it should help people learn faster, remember longer, discover connections earlier, and build deeper understanding across everything they know.

That future will require more than better prompts. It will require memory, routing, source awareness, local workflows, durable study systems, and interfaces that respect attention. Raxat lays more of that groundwork. It makes Zirn more configurable, more native, and more willing to preserve the work it has already done.

We're still at the beginning.

But each release brings us closer to software that doesn't simply store knowledge.

Software that helps cultivate it.

Welcome to Zirn 1.4 — Raxat.

The release that makes growing knowledge feel less like managing files and more like building understanding.

Zirn v1.4 is now available as a signed Mac download.