Grounded in your sources, or grounded in a falsifiable mechanism

Google renamed NotebookLM to Gemini Notebook in July 2026. Its strength is unchanged: answers stay inside the documents you give it. That is a different kind of grounding from the one here.

The short answer

Gemini Notebook answers only from what you uploaded. A model answers from a mechanism that states, in advance, the observation that would falsify it.

Both are attempts to stop an assistant from answering out of the undifferentiated everything. They constrain it in opposite directions.

Gemini Notebook constrains by corpus: it will not go outside your sources, and it cites which one each sentence came from. Excellent for "what do these forty documents actually say" — and it will not tell you the documents are wrong, because that is outside its remit.

A model constrains by mechanism: here is what drives outcomes, here is what would prove it wrong, here is when I retire it. It says nothing about your documents. It gives you a lens, and a way to find out later whether the lens was any good.

They compose. Ground a summary in your sources with one, then read that summary through a model with the other.

Side by side

Gemini Notebook (formerly NotebookLM)	Model Meets Reality
Grounded in your uploaded sources	Grounded in a stated mechanism
Citations back to a document	Falsifiers back to an observation
Answers what your corpus says	Answers what a mechanism predicts
Cannot tell you the sources are wrong	States in advance what would make it wrong
Your corpus stays in Google's cloud	Runs fully offline against a local model if you want

What Gemini Notebook (formerly NotebookLM) does better

Source grounding with per-sentence citation is genuinely hard and Gemini Notebook does it well, alongside audio overviews, mind maps and — since mid-2026 — running code against your material in a sandboxed notebook. If your question is "what is in these documents", nothing here competes. This site has no corpus ingestion at all.

Where this stands today

Said plainly, because the whole point of this site is not overclaiming. The Model Garden is opening small — no models listed as of September 2026. Every record shown is self-graded: the author's own count of claims made and resolved, labelled as such on each card. Nothing here is ranked, and no independent resolution layer exists yet. Gemini Notebook (formerly NotebookLM) has things this does not, named above rather than omitted. Compare the designs, not the scoreboards — there is no scoreboard.

https://github.com/someone/their-model Help me use this

Paste that into any assistant. It reads the model — premises, falsifiers, deletion clause — and reasons through it.

The difference that is not on the table

Gemini Notebook (formerly NotebookLM) answers your question. It does not keep a record of whether the answer held. That is not a flaw — it is not what an assistant is for. But if you have watched one thing closely for years, that knowledge is already a model: premises, predictions, blind spots. Writing it down turns it into something an assistant can reason through, and something a date can settle.

Share your understanding →

Browse the Model Garden Publish your own