What's New in Nimbus Studio: September 2026

The biggest change in August wasn't a feature — it was a shape change. Nimbus Studio stopped being only a pipeline builder and became a place to run apps and to create BCI projects with AI. Two surface additions lead this update — the Apps Hub and Create mode — and the rest of the work exists to make them trustworthy: a hardened calibration path, and the Nimbus Personalizer arriving as a first-class node. (If you missed the first wave of these changes, see Nimbus Studio & Nimbus SDK: June 2026 Product Updates.)

1. Apps Hub: one-click BCI experiences

The Apps Hub turns the Deploy runtime into a catalog of ready-to-run applications — no graph building, no setup:

• Phase 1 shipped the catalog + hub section, per-user profiles, and resume for interrupted sessions.
• Zero setup — sample datasets auto-download on first launch; the catalog filters by your hardware, with the full MI preset family included out of the box.
• Reliable replays — the demo execution cache is now versioned, so a stale pre-fix run can never replay forever.
• The Mind Reader (P300) demo app graduated into the complete calibrate → train → play journey — a second paradigm availabe to end users, not just researchers.

2. Create with AI mode: describe it, and it exists

Creating a BCI project is now a conversation with a canvas:

• Hub AI Create — a new Create shell that turns a prompt (or preset) into a project with a template-backed pipeline, now with an App Node sandbox preview so generated graphs can be inspected before you run them.
• Unified Project Canvas — projects move to four modes with shared data references, YAML docs, and the pipelines → projects merge: one canvas holds everything a project contains.
• Blank Create bootstrap — start from nothing without losing the guided flow, with unified Create/builder chrome across web and desktop.

3. Share and publish: BCI projects that travel

Create's header now carries the whole lifecycle — Run / Share / Publish — so work can leave your workspace:

• Share copies an authenticated Create deep link; anyone with project access opens the exact project and pipeline you're looking at. It builds on project sharing, live since May: invite by email, assign editor or viewer roles, and find shared work under Shared with me in the hub. Viewers get a read-only canvas; runs, artifacts, and visualizations are readable by the team while cancel, calibration, and deploy control stay owner-only — and execution ownership always stays with the runner.
• Duplicate, never collide — copy a project into your own workspace; nothing silently overwrites someone else's work.
• Publish points at the Apps Hub, where the first-party catalog is browsable today. User-published apps (a UGC gallery) come next — publishing is deliberately staged behind catalog quality.

4. P300: the second paradigm, end to end

P300 went from demo to a production journey in August:

• Row/column speller calibration engine with dedicated templates, a speller cue UI, and an explicit epoch capture window.
• Deploy correctness — streaming windows now match the trained epoch for xDAWN-flatten graphs (a silent accuracy tax, removed).

If you're using P300 in the lab, the end-to-end calibration flow is covered in BCI Calibration with Nimbus Studio: From Hardware to Trained Decoder.

5. Platform: the Personalizer node on nimbus-bci 0.6.0

• nimbus-bci 0.6.0 + personalizer node — the SDK's Nimbus Personalizer is now a Studio Model node: FEATURES-in, head (lda/qda/softmax), preset (strict/permissive/aggressive), adaptation (none / stream_head / gated_stream / affine / balora), plus a REVE→Personalizer template and full Adaptive Deploy + transparent Python export support.
• Desktop & assistant — if you haven't used it yet, Meet Your BCI Co-Pilot: The AI Assistant in Nimbus Studio is the best overview of what it can do on a real pipeline.

What's next

With apps and Create mode in place, the next chapter is letting users watch and govern adaptation — the uncertainty panel and decision-space work from July are the foundation, and the Personalizer's adaptive modes are the mechanism. For more on the SDK-side direction, see What's New in Nimbus: July 2026 SDK Update.