The conference in the car

Pill 2.8 · Series «The house notebook» (4 of 5) · A YouTube talk as a source and the Critique format of the Audio Overview KYBERVET Academy · Clínica El Roble · Phase 2: Make it work, not just answer · Video · 4:55

---

The talks from the last congress are on YouTube. They last forty minutes, they are in English and they lean on slides you can't see while you drive.

Gemini Notebook turns a talk into a two-voice audio for the car. If you don't tell it anything, it explains the talk so well that it sells it to you. In the Critique format, and with a brief, it separates what has been shown from what is a hypothesis, and leaves you with the questions you need to check. And you check them in the source, because the audio gets things wrong too.

At El Roble, the critique turned a fact upside down: where the speaker said hypercalcemia, the audio said hypocalcemia. Marta saw it on the slide.

Chapters · 4:55

1. 0:00 · Opening: the talk and the car
2. 0:24 · The talk comes in as text
3. 0:56 · Who organises it?
4. 1:17 · The default audio sells it to you
5. 1:51 · The critique, with a brief
6. 2:23 · What the critique says
7. 2:46 · In the car
8. 3:02 · The chat hasn't heard the audio
9. 3:49 · The slide at minute 29
10. 4:08 · What you decide

Read the full story

The scene

Monday, twenty to eight. The roundabout at the entrance to the city is like every Monday, and Marta has her phone in the dashboard holder with a talk in English playing through the speakers.

Amparo sent it last night, with one of her messages: “I want to cite it in my article. What do you think?”. It is from a 2024 congress: forty minutes on how chronic kidney disease starts and progresses in cats. Kidneys and physiology, Amparo's territory since before Marta started veterinary school.

Marta likes listening to these things in English. But a talk is also something you watch, and she is watching the road. The speaker says “as you can see on this graph” and Marta sees nothing. She says “hopefully you can appreciate here” and a bus cuts in on her right. Around minute eight, when she gets to the Starling forces in the glomerulus, the speaker warns: “this is where I might lose some of you, so stay with me”. Marta smiles. Amparo used to say the same in third year, with the same forces drawn on the blackboard.

And she gets lost all the same. Twice she has to rewind at a traffic light, and the second time she can no longer find where she was.

She parks on calle Olmos at twenty past eight. The sign is still lit and Sara is raising the shutter. Marta has the general idea and no opinion.

The push

That evening, at 20:10, she puts the talk into a new notebook and presses Audio Overview without changing anything. The next day she follows it from start to finish: two voices explaining it without needing a single slide, with examples about plumbing and offices. She arrives convinced.

With the engine already off, she stays in the car for a minute. The audio has ended by saying that one day the prevention of kidney disease will start “with cellular oxygenation therapies and lipid control from the cat's youth”. The speaker didn't say that. She presented the lipid idea as a possibility.

“They've told it better than she did,” she says out loud, alone in the car.

She texts Amparo what the talk says, in four lines. The reply arrives before Marta has crossed the pavement.

“That's what it says. I asked you what you think.”

The task

That afternoon, Marta adds to the notebook the two guidelines she opens whenever a cat comes in with high creatinine: the IRIS staging guidelines, in their 2026 version, and the ISFM consensus on feline chronic kidney disease. She uploads them as PDFs and clicks on each one, as Hugo taught her, to check they have come in complete.

Then she goes back to Audio Overview, and this time she doesn't leave it as it comes. She chooses Critique, in English, and writes a brief: to be critiqued by two vets from a neighbourhood clinic that sees a lot of older cats; to separate what has a study behind it from what is a hypothesis; to check it against the guidelines; to end with what changes anything in a practice like hers, and to stay away from drugs, doses and diets.

On Wednesday, in the car, the two voices pick at the talk's seams: that it mixes what has been proven with what is experimental without saying which is which, and that half a segment is molecular biology that can't be measured in a consulting room. It makes sense. But two sentences grate on her, and when she parks she writes them down on her phone. One, that the speaker presents lipids as a proven cause. The other, that on a diet very restricted in phosphorus the cats in one study developed hypocalcemia.

At 20:10 she gives them to the notebook's chat and asks, for each one, for the speaker's quote and what the guidelines say.

Both come back with a “not faithful”. On lipids, the speaker says there is no similar data in cats and calls it “a really interesting thought experiment”. On calcium, the talk speaks of hypercalcemia, total and ionized, in nineteen research colony cats. The hypocalcemia comes from the ISFM consensus, which says it is the most frequent finding in advanced disease. The audio had pinned it on the wrong cats.

Marta clicks on the calcium citation. The excerpt ends in “ionized hyp”, with the word cut off. She opens the video, takes it to minute twenty-nine, and there is the slide: nineteen cats, total calcium rising above the line and, at the top, the title of the original study. In the header, the name of the congress and of its organiser, a pet food brand.

Sara already has her scarf on. She stops behind her.

“And who pays for this?”

“She says so in the first ten seconds,” Marta replies, and takes the video back to the start so she can hear it.

Before closing, she writes a note in the notebook with what she decides: what she takes into the consulting room, what needs checking, what doesn't apply at El Roble and who organises the talk. And she replies to Amparo:

“Cite it, but present the lipid part as her hypothesis: she herself calls it a thought experiment. And cite the calcium part from the study on the slide.”

The reply takes a little longer than last time.

“And would that critique thing work for my article?”

“It would. It even thanks you for sending it in.”

“Better it fails me than the journal does.”

The unlock

If this is your first time opening Gemini Notebook, start with the video “What is Gemini Notebook” in pill 2.5.

One. A YouTube talk comes in as text. In “Add sources”, you paste the link to the video. What comes in is YouTube's automatic transcript, and nothing else. If the video has no captions, it doesn't come in. Whatever is only on the slides never reaches the notebook. The transcript has typos (“glamar filtration rate”) and the odd split word. It usually doesn't say which congress it was given at or who organises it either: that is in the video. And listen to the beginning, because many speakers declare there who funds them.

Two. The format decides what you take to the car. When you press Audio Overview, “Customize Audio Overview” opens. If you generate it as it comes, you get “Deep Dive”: two voices that explain it very well and that, to make it easy to follow, round things off and even add. In our test, the audio ended with a promise the speaker doesn't make.

The “Critique” format describes itself as an expert review of your sources, with constructive feedback to help you improve your material. It evaluates: what has been shown, what is a hypothesis and what is missing. The brief goes in the box “What should the AI hosts focus on in this episode?” and decides three things:

• where the critique comes from: a neighbourhood clinic with a lot of older cats;
• what it is checked against: the guidelines you have added alongside;
• what stays out: drugs, doses and diets.

You choose the language. The length can be “Shorter” or “Default”; the “Longer” option only appears in English and with “Deep Dive”. In our test, the critique in English came out at sixteen minutes, and the one in Spanish at twelve.

Three. Whatever makes you doubt is checked in the source. The critique leaves you with questions, but it isn't the truth. In our test it was unfair to the speaker in two places. And, with the guidelines alongside, it pinned a fact from a guideline on the talk.

The notebook's chat hasn't heard the audio: if you ask it what it says, it imagines it. So you write down the doubts yourself and give them to the chat, so it gives you the quote from the talk and from the guidelines. Click on the citation. If it is cut short, go to the video, to that minute, and look at the slide.

To cite something in an article, a session or the morning briefing, use the talk, the guideline or the original study. Never the audio.

What came out of El Roble

The talk: “Pathophysiology of Early and Progressive CKD in Cats”, 39 minutes in English, from the Companion Animal Nutrition Summit 2024, organised by the Purina Institute. Alongside it, two guidelines in PDF: IRIS Staging of CKD (2026) and the ISFM consensus on feline chronic kidney disease (2016).

	Default	Critique, talk only	Critique, with the guidelines
Language and length	Spanish, 13:35	Spanish, 11:55	English, 15:57
What it does well	Explains the pathophysiology very clearly	Separates what has been proven (the 213-cat study) from what is experimental	Checks against the guidelines: ionized calcium, blood pressure, proteinuria
What it misses	Treats the lipid idea as almost settled and ends with a promise the speaker doesn't make	Says the speaker acknowledges her limits “on the last slide”. She does so halfway through, and in the first few seconds she already declares who funds her	Says she presents lipids as a proven cause and swaps hypercalcemia for hypocalcemia

None of the three mentions the first thing the speaker says: that she has received funding from industry and from foundations.

The two doubts, checked in the chat:

What the critique said	What the talk says	What the guidelines say
That she presents lipids as a proven cause	“while similar data are not available for cats…”, and she calls it “a really interesting thought experiment”	They don't cover it
That on a diet very restricted in phosphorus the cats developed hypocalcemia	Hypercalcemia, total and ionized, in 19 research colony cats (Schauf et al., J Vet Intern Med, 2021)	The ISFM says that in advanced disease ionized hypocalcemia is the most frequent finding. Hence the mix-up

The critique tells you what to check. What you take into the consulting room is your decision.

Do it yourself in 4 steps

1. Prepare the notebook. In “Add sources”, paste the YouTube link to the talk. Add the reference guidelines on the topic alongside it, in PDF, and click on each source to check it has come in complete.
2. Get the critique. In Studio, press “Audio Overview” and “Customize Audio Overview” opens. Choose “Critique”, the language and the length. In “Sources”, keep the talk and the guidelines. Paste the brief into “What should the AI hosts focus on in this episode?” and press “Generate”.
3. Listen to it in the car. Download the audio in the phone app: it plays offline and with the screen locked. Whatever makes you doubt, write it down when you park.
4. Check and decide. Paste the checking question with your doubts into the chat. Click on the citations and, if you need to, look at the slide in the video. Write a note with what you decide.

The critique brief · copy it

You are two vets from a small neighbourhood clinic that sees a lot of older cats.
Critique the talk. Don't summarise it.
Check the important points against the guidelines in the notebook:
whether they agree, disagree or don't cover it. The guidelines are not to be critiqued.
Separate what the speaker supports with studies from what is hypothesis or opinion.
End with three things: what changes anything in our practice, what doesn't, and what should be checked.
No specific drugs, doses or diets.

Change the first line to describe your clinic and your patients.

The checking question · copy it into the chat

I want to check two things the critique says about the talk.
For each one, tell me:
1. What exactly the speaker says, with the quote.
2. Whether she supports it with a study, which one and with how many animals.
3. What the guidelines in the notebook say, with the quote. If they don't cover it, write “not stated”.
4. In one line: whether the critique is faithful to the talk (yes, no or partly).
Don't give me treatment recommendations.
First: (what made you doubt)
Second: (what made you doubt)

The note of what you decide · copy it

I take into the consulting room:
Needs checking:
Not in my clinic:
Who organises it and who pays:

Before → Now

Before. The talk waiting in “Watch later” since spring. Or played as it is in the car, to arrive with the general idea. Or a summary that sells it to you.

Now. Fifteen minutes of critique in the car, two doubts written down when you park, ten minutes with the chat and the slide, and a note with what you decide. What you cite comes from the source.

Before you start

• No captions, no source. The video has to be public and have a transcript. If it is a livestream that has just ended, sometimes it doesn't have one yet.
• What is on the slides isn't there. Only what is said comes in. If a figure only appears on a slide, the notebook doesn't know it.
• The congress and the sponsor, in the video. The notebook usually doesn't know them. Look at the header of the slides and listen to the first few seconds.
• “Longer”, only in English and with Deep Dive. The Critique only has “Shorter” and “Default”. None of the audios in our test went over sixteen minutes.
• The same format can have two names. In the Spanish interface, Deep Dive is called «Información detallada» in the customize window and «En profundidad» in the Studio list.
• Guidelines alongside help, but they can mix up facts. In our test, the audio pinned a fact from the guideline on the talk.
• The chat hasn't heard the audio. If you ask it what it says, it imagines it. You write the doubts down yourself.
• The Critique talks as if the material were yours. At the end it thanks the author for sending it in. That is why it also works for what you write yourself.
• Drugs, doses and diets, out of the brief. Even so, some slip through: in our test, the English audio talked about when to “intervene immediately”. Clinical matters are your decision.
• It takes time and uses up your quota. A default audio took about 7 minutes, and the long critique with guidelines about 15. With the free account, two audios took usage to 80%. At El Roble they do it with the clinic's paid account.
• At the wheel, just listen. Doubts are written down when you park.
• The buttons will keep changing. What will stay the same: having the talk critiqued instead of summarised, checking in the source whatever makes you doubt, and writing down yourself what you decide.

Support material

The guide, to keep at hand. The four audios from our test, so you can hear the difference between the summary and the critique. And the sources, which are what you cite.

The four audios from the test

These are the audios Gemini Notebook generated in our test, exactly as they came out. An AI made them and they contain the errors the pill describes: that is why they are here. Listen to compare, and don't cite them. Three are in Spanish and one in English. Each comes with its automatic transcript, checked by hand only for names and technical terms.

The sources, first-hand

To cite something in an article, a session or the morning briefing, go here. Each link takes you to the publisher's website.

• Pathophysiology of Early and Progressive CKD in Cats The talk, on YouTube: 39 minutes in English. Companion Animal Nutrition Summit 2024, organised by the Purina Institute.
• IRIS Staging of CKD (2026) The IRIS staging guidelines, on the IRIS website.
• ISFM Consensus Guidelines on the Diagnosis and Management of Feline Chronic Kidney Disease Sparkes et al. J Feline Med Surg, 2016.
• Schauf et al. J Vet Intern Med, 2021 The study on the slide at minute 29: the 19 research-colony cats.

---

Next week · 2.9 · The protocol the locum sees in two minutes. Diego has the clinic's protocols in a document nobody reads all the way through. The notebook makes him a short video for whoever comes in to cover.