Changelog
Everything that ships, in the order it shipped. Newest first.
On the model behind P3ZO. We'll be straightforward about this: replies are currently generated by LLM7. It's a capable general model and it's what the preview runs on today. As our dataset grows we intend to move to our own model, used alongside a selection of the strongest models available, and route each request to whichever is best suited to it. That work is underway rather than finished — this note will change when it does.
0.4.0NewLatest
- Local chat history. Conversations are saved in your browser and grouped by date in the sidebar.
- Landing page. A single-screen entry point with an ask bar that carries your question straight into the chat.
- Blog and changelog. The pages you're reading now.
- Model transparency. The provider running the preview is now named openly rather than left unsaid.
NextPlanned
- Our own model. Trained on the dataset we're building, and introduced gradually rather than in one switch.
- Multi-model routing. Combining our model with other leading models, choosing per request instead of relying on a single provider.
- No date attached. This is direction, not a commitment — we'd rather say so than quote a timeline we can't hold.
0.3.0Improved
- Refined interface. Softer surfaces, larger type, and a calmer palette throughout.
- Simplified setup. Model selection removed — there is nothing to configure before you start.
- Copy buttons. Individual replies and code blocks can be copied in one click.
0.2.0Fixed
- Clearer errors. Failed requests now explain what went wrong instead of showing a raw status code.
- Interrupted replies. Stopping mid-response keeps whatever had already arrived.
- Mobile layout. The composer no longer collides with the on-screen keyboard.
0.1.0Initial
- First build. Streaming replies, markdown rendering, and a single-page client.