Gemini: the complete course

1. What Gemini is (and why it has so many names)

Gemini is a brand, not a single program: like Fiat, which makes the Panda but also the delivery van.

If you have tried to understand Gemini by reading the news, you have probably got confused. One day it is «Google's chatbot», the next it is «a model», then «an app», then «a subscription». They are all right, because the word is used for three different things.

• The models: the real artificial «brains», trained by Google (for example Flash and Pro).
• The app and the website: the window where you talk to those brains, at gemini.google.com and in the Windows app.
• The tools built around them: Gemini Notebook for studying, Antigravity to let an agent work on your computer, the open Gemma models.

In this course we will see every piece, one at a time. Today we tidy the garage: which engines exist, how much they cost and which one you really need.

2. The models: light, balanced, powerful

When you open Gemini on the web, next to the writing box you find a word, for example Flash, with a small arrow. That is the model selector. Click it and a small menu opens: the list of engines available on your account.

SCREENSHOT

Drop-down menu of models: 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and Extended reasoning

The model selector in Gemini on the web (7 October 2026, Google AI Pro plan). The interface is in Italian.

On my account, that day, the list said this. Each entry has a name and a short description written by Google.

2.1 The number does not tell you who is best

Look closely: there is a 3.8 Flash and a 3.1 Pro. A higher number does not mean «more powerful»: it shows the generation. The name (Flash-Lite, Flash, Pro) shows the role. A new Flash may be more up to date than an older Pro, but it is still built to be nimble, not for the longest reasoning.

Diagram

Four roles, four choices

Four roles, four choices

3. The plans: free, Plus, Pro and Ultra

Gemini can be used for free, but paid plans widen the limits and add features. The simplest way to see them is to think of a gym membership: with the free one you get in and use the basic machines; with the pricier ones you get more visits, more classes and the weights room.

3.1 And if it is a company account?

If you use Gmail and Google Drive with a work account (Google Workspace), the administrator decides what is switched on. In that case the privacy and data rules change: ask whoever manages the account before pasting confidential documents.

4. Usage limits: the fuel tank

Every plan has a fuel tank. The more you ask, the emptier it gets; then it refills by itself. Gemini shows it in plain sight: from the settings menu (the cog at the bottom left) choose Limiti di utilizzo (Usage limits).

SCREENSHOT

Gemini Usage limits page showing the current usage and weekly limit bars

The usage limits page: current usage, with the refill time, and weekly limit.

The page has two bars. Current usage is the short tank, with the time it refills; Weekly limit is the long one, with the reset day. The day I took the screenshot I was at one per cent and three per cent: I had just started.

• Long requests and the most powerful models use more than a short question to Flash.
• Images, video and Deep Research weigh more than a normal chat.
• If you always run out of tank, that is the signal to consider a higher plan, not before.

5. How to choose without going crazy

Here is a practical rule: start from the free plan and change only when you feel the need. A three-by-two-metre pond does not need an Olympic swimming pool pump.

1. Use Flash to write, summarise, translate and answer everyday questions.
2. Move to Pro when the task has many pieces to hold together (a plan, an analysis, a long text).
3. Try extended reasoning when Pro's answer feels hasty on a really complicated problem.
4. Consider Plus or Pro when you often hit the limits, need video or want more room in your Google Drive.
5. Skip Ultra unless AI is your daily work tool.

Hands-on exercise

Fifteen minutes, a Google account and the website gemini.google.com.

1. Open gemini.google.com and sign in. Click the model name next to the writing box and note the entries you see, with their description.
2. Write the same question with two different models, for example Flash and Pro: «Explain to a child why the water in a pond turns green in spring».
3. Compare the two answers: length, clarity, time taken. Which one do you prefer for this task?
4. Open the settings menu, go into Usage limits and note the current and weekly usage percentage.
5. Write in one line which plan you have (or which one seems right for you) and why.

Questions that come up in class

Is Gemini free?

Yes, there is a free plan with Gemini on the web and in the app. Paid plans raise the limits and add features, such as video generation.

What is the difference between Gemini and Google AI Pro?

Gemini is the artificial intelligence; Google AI Pro is the subscription that gives you higher limits, more storage and the most powerful model. It is like the difference between the engine and the gym membership.

Why is 3.8 Flash newer than 3.1 Pro?

Because the number shows the generation, not the power. Flash and Pro are two different roles: a Flash version may come out after a Pro version.

Can I change model in the middle of a conversation?

Usually yes: the selector is always next to the writing box and applies to the following requests. If the answer changes a lot, remember that the new model may read the earlier conversation differently.

Is Gemini Notebook the same as Gemini?

No: it is the new name of NotebookLM, a tool for studying your own documents. We look at it in lesson 9. From the Gemini app, however, you open it from the sidebar.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, Google AI plans (gemini.google/subscriptions)

Italian page checked on 7 October 2026

Gemini app on the web, model selector and Usage limits page

author's screenshots, 7 October 2026

1. The front door

To start you do not need to install anything: a browser and a Google account are enough.

Open your browser and go to gemini.google.com. If you already have Gmail or an Android phone you are all set: you sign in with the same account. From then on all conversations are tied to the account, not to the computer. If tomorrow you open Gemini on another PC, or in the Windows app we cover in lesson 8, you find the same chats.

SCREENSHOT

Gemini home with sidebar, writing box and model selector

Gemini's start screen on the web. Recent chats in the sidebar have been blurred. The interface is in Italian.

Here is what you see, from left to right.

• Sidebar: Nuova chat (New chat), Cerca nelle chat (Search chats), Video, Raccolta (Library), the Notebook section, then the list of recent chats.
• Attività (Activity) and account at the bottom of the bar, with the settings cog.
• Writing box in the middle, with the + button to attach files and choose tools, the model selector and the microphone.
• The greeting, which changes every time («E ora, Gianluca?», «Cominciamo»...). It means nothing: it is just courtesy.

2. The first conversation

Let's try it with a real case. Imagine you want to dig a small garden pond of three metres by two, and you do not know where to start. Click in the box and write your request in your own words. I wrote this (in Italian, my language):

Press Enter and wait a few seconds. Gemini writes the answer as it goes and gives the chat a title (here «Come realizzare un laghetto da giardino», How to build a garden pond), which immediately appears among the recent chats.

SCREENSHOT

Chat with an answer of five numbered steps about digging a pond

The first answer: five numbered steps, from marking out the dig to laying the liner.

Notice the form: a numbered list, key words in bold, measurements (steps at 20-30 centimetres, 40-50, and a central hole of at least 80). Why? Because I asked for «five steps» and «in a simple way». You decide the shape of the answer with your request.

2.1 Continuing the conversation

The nice part is that Gemini remembers what you said in that chat. You can write «Explain point 3 better» or «Rewrite it for a ten-year-old» without repeating everything. If you change subject, open a New chat: one subject per chat keeps everything tidier and more precise.

3. The icons under the answer

At the bottom of every answer there is a row of small icons. They are few, but they save a lot of time.

SCREENSHOT

Row of icons under the answer: thumbs up, thumbs down, regenerate, copy, more

The icons under the answer: thumbs up, thumbs down, regenerate, copy and more actions.

At the top right you also find the Converti chat in PDF (Convert chat to PDF) button: it turns the whole conversation into a document, handy to keep a guide or send it to a friend.

4. Finding and organising chats

After a few weeks you have dozens of chats. To keep track you have three tools. The first is Cerca nelle chat (Search chats), in the sidebar: type a word and Gemini searches your conversations.

The other two are in the three-dot menu at the top right, when a chat is open.

SCREENSHOT

Menu with Share conversation, Pin, Rename, Delete

The chat menu: Condividi conversazione (Share), Fissa (Pin), Rinomina (Rename), Elimina (Delete).

• Pin: the chat stays at the top of the list, like a sticky note on the fridge. Ideal for guides you consult often.
• Rename: Gemini picks a title, but you can give your own, clearer one.
• Share conversation: creates a link you can send to someone. Anyone with the link can read.
• Delete: removes the chat. Think before confirming: do not count on getting it back.

5. The settings menu, entry by entry

At the bottom left, next to your name, there is a cog. Click it: a long menu opens. It looks like a jungle, but every entry has a precise place in the course.

SCREENSHOT

Gemini settings menu with all entries

Gemini's settings menu, with the main entries.

6. Privacy: what to write and what not

Everything you write leaves your computer and travels to Google's servers. That is normal, it is how it works, but it asks for a little common sense. The rule I give my students is the sign on the fence: if you would not hang it outside your house, do not paste it into the prompt.

• No passwords, codes, card numbers or identity documents.
• No confidential data of customers or colleagues, unless you are authorised.
• Caution with medical and legal information about other people.

At the bottom of the sidebar there is the Attività (Activity) entry: it takes you to the page where Google shows the history of your requests and the settings on what is saved and for how long. Open it once, calmly, and decide for yourself. Your own data appears there, so do not share screenshots of it.

Hands-on exercise

Twenty minutes to go all the way round with a test conversation.

1. Open gemini.google.com and sign in with your Google account.
2. Write a request about a real project of yours, with the number of points and the tone you want. For example: «Give me 5 steps to prepare the vegetable garden in spring, in a simple way».
3. Ask a follow-up about the point that interests you most, without repeating the context.
4. Click the circular arrow to regenerate the answer and compare the two versions.
5. Rename the chat with a title of your own and pin it at the top of the list.
6. Open the settings menu and look at I tuoi link pubblici and Attività, without changing anything.

Questions that come up in class

Do I absolutely need a Google account?

Yes, as a rule you need a Google account to use Gemini and save chats. A personal one is fine: you do not need Workspace.

Can I find my chats on another computer?

Yes: conversations are tied to the account. I made them on the web and found them identical in the Windows app.

Can I delete a conversation?

Yes, from the chat's three-dot menu with the Delete entry. To delete in bulk or decide how long things are kept, go through the Activity page.

Why does Gemini say wrong things so confidently?

Because it builds the most probable text; it does not check facts like an encyclopaedist. For measurements, prices and regulations, always verify.

Can I share a chat with a friend?

Yes, with Share conversation: you get a link. Anyone who has it can read it, so do not use it for chats with personal data.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Gemini on the web (gemini.google.com)

author's screenshots, 7 October 2026

Google, privacy notice for the Gemini apps

read the up-to-date version for your country

1. Why the answer is vague

Gemini cannot read your mind: it works only with what you write.

Imagine walking into a trattoria and telling the waiter: «Bring me something good». You get the most common dish, which is almost never the one you had in mind. With Gemini the same thing happens: if you write «tell me about ponds», you get a generic summary, because the system has to guess who you are, what you need and how long you want it.

The good news is that writing a good request takes just a four-step method, which you learn in ten minutes.

2. The four steps: context, goal, format, verification

Diagram

The four steps of an effective prompt

The four steps of an effective prompt

Notice that the «after» is not harder: it is just more complete. And the fourth step, verification, is the one almost everybody skips. A text that sounds good is not necessarily true: asking Gemini to state its assumptions turns a self-confident answer into a checkable one.

3. Attaching files, photos and documents

Sometimes the context does not fit in two lines: it is a photo of the garden, a quote in PDF, a table. That is why next to the writing box there is the + button.

SCREENSHOT

Menu with Upload file, Add from Drive, More uploads, Create image, Create video, Create music, Canvas and More tools

The + button menu: attachments at the top, creation tools below. Interface in Italian.

The menu has two halves. At the top are the ways to bring in your materials; below, the tools to create something (images, video, music, Canvas), which we see in lessons 4 and 5.

• Carica file (Upload file): pick a file from your computer, for example a PDF, a photo or a document.
• Aggiungi da Drive (Add from Drive): take a file straight from your Google Drive, without downloading it.
• Altri caricamenti (More uploads): other sources to take material from.

3.1 What you can ask about an attachment

1. «I am attaching a photo of the garden: in which corner does the pond make most sense and why?»
2. «Summarise this quote in five lines and tell me which items seem the most expensive.»
3. «This table is messy: rewrite it cleanly, with one column per material.»

4. Choosing the right model for the request

In lesson 1 we saw the models. Now we use them. A practical rule: do not send the truck to buy bread.

5. Conversing: the request improves as you go

Nobody writes the perfect request first time, not even people who do this for a living. The trick is to treat the first answer as a draft. Then you work with follow-up questions. Here are five moves I use all the time.

1. Get it to ask you. Write: «Before answering, ask me three questions you need to understand my case better.» Gemini interviews you and the final answer is much more targeted.
2. Ask for alternatives. «Give me three different versions, a shorter one, a more technical one, a funnier one.»
3. Show an example. «Write in the style of this text» and paste a paragraph of yours.
4. Narrow down. «Now cut it in half and remove every difficult word.»
5. Ask what it does not know. «Which of this information can you not know for certain?»

Remember also the circular arrow from lesson 2: if an answer does not convince you, sometimes it is enough to have it regenerated. But if it keeps being wrong, do not insist: change the request.

6. The fourth step: verifying

Gemini is brilliant at writing sentences that sound right. That is its strength and its risk: sometimes it invents a fact with the same confidence with which it states a true one. Experts call it a hallucination.

• Always ask for the assumptions and the things to check.
• For numbers, measurements, prices, dates and regulations, find a second source (an official site, a book, an expert).
• If the answer concerns health, money or law, treat Gemini as a first draft, never as a decision.
• Ask it to cite sources, then open them yourself: a link that does not exist is an alarm bell.

Hands-on exercise

Twenty minutes, a real task from your work or home and, if you like, a file to attach.

1. Choose a task you do often, for example answering an email, preparing a quote or organising a job in the garden.
2. Write the request the way you would today, in one line. Save the answer.
3. Rewrite it with the four steps: context, goal, format, verification. Compare the two answers.
4. Attach a file or a photo with the + button (even a photo taken with your phone is fine) and ask for a summary or an opinion.
5. Ask: «Before answering, ask me three questions» and reply. Notice how much the answer changes.
6. Ask for the assumptions made and check at least one fact on a second source.

Questions that come up in class

Does a longer prompt work better?

No: precision matters. Five lines with the right information beat a page of waffle.

What kind of files can I attach?

Documents, PDFs, photos and more: the + button menu shows the options on your account. Supported formats change over time and with the plan.

Does Gemini keep the files I attach?

Files become part of the conversation. To learn how they are handled and for how long, read the privacy notice for the Gemini apps and the Activity page.

Can I use Gemini in Italian and English together?

Yes, you can write in one language and ask for the answer in another. For important texts, have a person read them.

What if the answer still does not convince me?

Almost always the format is missing, or written vaguely. Add an example of the result you want: it is the most powerful technique.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Gemini on the web, + button menu and model selector

author's screenshots, 7 October 2026

Google, guide to using Gemini and privacy notice for the Gemini apps

to be consulted in the up-to-date version

1. Where the tools are

A normal chat is a conversation. The tools are different instruments, one for each job.

Click the + button next to the writing box and then Altri strumenti (More tools): a second menu opens with Canvas, Deep Research and Apprendimento guidato (Guided learning).

SCREENSHOT

Plus button menu with the More tools submenu: Canvas, Deep Research, Guided learning

The More tools menu with Canvas, Deep Research and Guided learning. Interface in Italian.

When you pick a tool, a label appears in the box (for example Deep Research) and stays until you remove it. To go back to the normal chat just click the small cross.

2. Deep Research: the researcher who works for you

When you ask a normal question, Gemini answers in a few seconds with what it already knows. With Deep Research something else happens: Gemini prepares a research plan, reads many web pages, compares the information and in the end writes a complete report. Imagine sending a tireless assistant to the library and finding on your desk a write-up with the sources.

SCREENSHOT

Writing box with the Deep Research label and the Sources and File buttons

The Deep Research box: «Che cosa vuoi cercare?» (What do you want to search?), with the Fonti (Sources) and File buttons.

2.1 How to use it, step by step

1. From the tools menu choose Deep Research. The box changes and asks: «What do you want to search?».
2. With the Fonti (Sources) button choose where to search. Google says that by default there is web search, and apps such as Gmail, Drive and Notebook can be added. With File you can attach your own documents.
3. Write the question, with context and goal (use the four steps from lesson 3).
4. Gemini shows you a plan: you can edit it, then you start the research.
5. Wait. According to Google it usually takes five to ten minutes, more for complex topics. You can leave the chat and come back later.
6. Read the report, open the sources and, if you like, export it to Docs.

And remember the fourth step of the method: verify. The report looks like a degree thesis, but it is still a text written by an AI. Open at least two of the cited sources and check that they really say what the report claims.

3. Guided Learning: the teacher, not the encyclopaedia

There is a big difference between getting an answer and learning. An encyclopaedia gives you the answer; a good teacher asks questions, starts from what you know and takes you along step by step. Guided learning is the second thing.

SCREENSHOT

Guided learning answer about the nitrogen cycle with three numbered paths

Guided learning: instead of dumping everything, Gemini proposes three paths and asks where to start.

I wrote: «Teach me how the nitrogen cycle works in a pond, starting from zero». Gemini did not pour a lecture over me. It explained in a few lines why the problem exists (fish, fallen leaves and uneaten food release substances that become dangerous) and then asked where I wanted to start, offering three numbered paths: the chemical sequence and the bacteria, starting up and managing the filter, the plants and closing the cycle.

• In the box appears the label Apprendimento (Learning) and the question «Che cosa vuoi imparare?» (What do you want to learn?).
• Answer its questions: that is how it understands your level.
• You can ask at any moment: «Quiz me» or «Ask me three questions to see if I understood».
• It works for school subjects, for a hobby, for a new program to learn.

4. Canvas: the workbench next to the chat

Have you ever tried to polish a long text inside a chat? Every correction is a new message, and the text gets lost among the others. Canvas solves the problem: it opens a worksheet next to the conversation, where the document stays still and you edit it calmly.

SCREENSHOT

Canvas with the document Checklist Laghetto da Giardino 3x2 m

Canvas with the checklist to build a 3-by-2-metre pond, conversation on the left.

I chose Canvas and asked: «Write a one-page checklist to build a 3-by-2-metre garden pond, with items to tick». About ten seconds later the document was ready, with a title, a short introduction and numbered sections (planning and positioning, then the following phases).

4.1 What you can do in the sheet

• Edit the text by hand, as in an editor, or ask Gemini to change a piece (for example: «Make section 2 shorter»).
• Undo and redo changes with the arrows at the top.
• Enlarge the sheet to full screen or share it.
• Export the work with the Esporta (Export) button.

SCREENSHOT

Canvas Export menu with Docs and .md

The Export button: the two choices are Documenti (Google Docs) and .md.

The two routes are Documenti, that is Google Docs in your Drive, handy if you then work with Word or with other people, and .md, a text file with simple formatting, which you see in programs such as advanced notepads. For a document to print or send by email, the first is almost always the right choice.

5. Which tool for which job

In practice a complete job can use them all, in sequence. For Marta's pond: Deep Research to choose the plants, Guided learning to understand the filter, Canvas to write the checklist and take it to Docs.

Hands-on exercise

Twenty-five minutes, three tools, one project.

1. Choose a topic you really care about (a hobby, a job, a trip).
2. Open Deep Research from the More tools menu and ask a question with context and goal. Let it work and, when it is ready, open two sources and check one fact.
3. Open a new chat with Guided learning and ask to be taught something on the same topic starting from zero. Answer its questions and at the end ask it to quiz you.
4. Open Canvas and ask for a checklist or a one-page sheet on the topic.
5. Edit one item by hand and ask Gemini to shorten a section.
6. Export the document to Docs and open it in your Drive.

Questions that come up in class

How long does Deep Research take?

According to Google, usually five to ten minutes, more for complicated topics. You can leave the chat and come back later.

Can I trust the Deep Research report?

It is an excellent starting point, not a verdict. Open the cited sources and check the important data, especially numbers, regulations and prices.

Does Deep Research use much of my limits?

Yes, far more than a normal question. If you have a free plan, keep it for the research that really matters.

Is Guided learning a different model?

No: it is a mode, a different way to run the conversation, which uses Gemini's models.

Where does the document I create in Canvas end up?

It stays in the conversation. To have it as a file, use Export and choose Docs or .md.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, How Deep Research works in the Gemini app

help page consulted on 7 October 2026

Gemini on the web, tools menu, Canvas and Guided learning

author's tests and screenshots, 7 October 2026

1. Three creative studios in the same place

Images, video and music: the right words are enough, like when you tell an idea to an artist friend.

In the + menu of the writing box, below the row of attachments, you find three entries: Crea immagine (Create image), Crea video (Create video), Crea musica (Create music). Each opens a page with ready-made examples and a box where you describe what you want. In the sidebar, under Search chats, there is also the Video entry, a direct access to the Create video page.

2. Images: the recipe with five ingredients

Open Create image. Above the box you find many sample thumbnails (80s styles, monochrome, runway and others): click one and Gemini uses it as a starting point. If you start from zero instead, describe in your own words.

SCREENSHOT

Gemini Create images page with sample styles and the Aspect ratio button

The Create images page, with sample styles and the Proporzioni (Aspect ratio) button. Interface in Italian.

A good description is like a recipe. There are five ingredients: subject, setting, style, light and framing. You do not need all of them every time, but the more you add, the less Gemini has to guess.

Diagram

The five ingredients of an image description

The five ingredients of an image description

I wrote a single line (in Italian): «A garden pond with pink water lilies at sunset, watercolour style, warm light». About twenty seconds later this image appeared.

SCREENSHOT

Chat with a watercolour of a pond with pink water lilies at sunset created by Gemini

The image created by Gemini from the description above. Icons to share, copy and download appear on the image.

2.1 Polishing the result

• Continue in the same chat: «Make the sky more purple», «Add a frog on a leaf», «Remove the people».
• With the Proporzioni (Aspect ratio) button choose the shape of the image (for example wider for a banner, taller for a story).
• You can upload a photo of your own with the + button and ask for edits: change the background, remove an object, turn it into a drawing.
• If you do not like it, do not insist on the same sentence: change the style or the light.

3. Video: a few seconds, many decisions

The Create video page works like the images one: ready examples (such as Tiny world, Anime, 8-bit adventure, Logo reveal, Product showcase) and a box where you describe the scene. Below the box are the options, for example orientation: Orizzontale (Horizontal) for YouTube, or vertical format for social media.

SCREENSHOT

Gemini Create video page with examples and the Horizontal option

The Create video page, with the examples and the orientation option.

A video is an image that moves, so to the image description you add movement and rhythm: «the camera slowly lowers toward the water», «a dragonfly crosses the frame from left to right».

4. Music: from a hum to a soundtrack

Create music is the surprise of the family. You describe a track in words and Gemini composes it. The page offers examples (background music, intro music, brand jingle, soundscape) and, below the box, three options to adjust.

SCREENSHOT

Gemini Create music page with the options Length, Vocals and Genre

The Create music page, with the options Durata (Length), Parti vocali (Vocals) and Genere (Genre).

• Length: how long the track should last.
• Vocals: with or without a voice.
• Genre: the musical style.

To describe music think of three things: mood («relaxing, like a garden in the morning»), instruments («acoustic guitar and a bit of piano») and what it is for («for a one-minute video about my pond»). A ten-second jingle for your channel is a great first experiment.

5. Using them well: people, rights and watermark

Creating is easy, but using things with judgement is your job. Here are three recommendations, from an older brother.

1. Real people. Do not ask for images or videos of real people (colleagues, relatives, public figures) without their permission. Fake images of real people can do damage, even unintentionally.
2. Rights and trademarks. Do not ask to copy an artist's work or a company's logos. If you use the results for work, read Google's terms of service on commercial use.
3. Say it is made with AI. If you publish an image created by Gemini, say so. In the settings menu there is the entry Filigrana dei contenuti multimediali (Media watermark), which explains how Google marks content created with AI.

Hands-on exercise

Twenty minutes, the five-ingredient recipe and a project of yours.

1. Choose a subject you love (your garden, a dish, your shop) and write the description with the five ingredients.
2. Open Create image from the + menu and paste the description. Look at the result and change only the light or the style, then compare.
3. Try changing the aspect ratio with the Proporzioni button for a banner version.
4. Upload a photo of yours and ask for a simple edit (for example, removing an object from the background).
5. Open Create music and describe a jingle of a few seconds for your project: mood, instruments, what it is for.
6. Before publishing anything, reread the three recommendations: real people, rights, disclosure.

Questions that come up in class

Can I use images created by Gemini for work?

It depends on Google's terms of service and on your case. For commercial use (shop, advertising, books) read the up-to-date terms and, in important cases, ask a professional.

Why does the image sometimes have a wrong detail?

Because generators struggle with hands, long texts and very small details. Ask for a precise correction in the same chat, or regenerate.

How much do images, video and music use?

More than a normal question, and videos far more than images. The Usage limits page (lesson 1) shows you the tank.

Can I edit a photo of mine?

Yes: you upload it with the + button and describe the edit. Do not upload photos of other people without their consent.

Are the created contents recognisable as AI-generated?

Google states that it marks content with a watermark. For the details open the Media watermark entry in the settings.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Gemini on the web, Create images, Create video and Create music pages

author's screenshots and tests, 7 October 2026

Google, The Gemini app is now available for Windows (blog)

for the statement that some video features require a Google AI subscription

1. Three ways to be understood better

Repeating the same things to Gemini every day is like explaining to the barman again that you want your coffee with a drop of milk. Better if it already knows.

Gemini has three levels of personalisation, each with a different job. It is worth understanding them before using them, because something changes in November.

Diagram

The three levels of personalisation

The three levels of personalisation

2. Personal context: memory and instructions

From the settings menu choose Contesto personale (Personal context). The page has two blocks, each with its own switch.

SCREENSHOT

Personal context page with Memory and Your instructions for Gemini

The Personal context page: Memoria (Memory) and Le tue istruzioni per Gemini (Your instructions for Gemini). Interface in Italian.

• Memory: Gemini learns from your previous chats to get to know you better. With the Gestisci ed elimina (Manage and delete) link you can review and delete past conversations.
• Your instructions for Gemini: fixed rules you want to apply always. For example, the page itself suggests: «Start answers with a TL;DR summary» or «Use bullet lists for long paragraphs». The Aggiungi (Add) button is for writing new ones.

3. Gems: tailor-made assistants

A Gem is a version of Gemini with ready instructions for a task. Instead of writing every time «you are a nurseryman, answer with numbered steps...», you write it once and then call the Gem.

SCREENSHOT

Gem manager with the 17 November notice, default Gems and My Gems

The Gem manager, with the notice about the move to skills, Google's default Gems and my Gems.

In the Gem manager you find two sections: the Google defaults (Storybook, Idea generator, Career advisor, Programming partner and others) and My Gems. To create one click Nuovo Gem (New Gem).

SCREENSHOT

Editor of the Gem Consulente Laghetto with name, description, instructions and preview

A Gem's editor: name, description, instructions and, on the right, the preview.

3.1 The editor fields

• Name and Description: they help you recognise it.
• Instructions: the heart of the Gem. Under the box there are undo and redo buttons and a button with a wand, which helps you rework the text.
• Default tool: you can attach a tool (for example images or Deep Research) or leave «No default tool».
• Knowledge: files the Gem can refer to, such as a manual of yours.
• Preview: on the right a small chat to test.

3.2 How to write good instructions

The best instructions have four parts, which you can remember as who, what, how, in what form. This is the Gem I wrote for the course (in Italian, translated here):

4. From 17 November Gems become skills

At the top of the Gem manager, on 7 October 2026 (the day I checked), there is a notice worth reading calmly.

What does it mean? That the instructions you wrote in your Gems will not be lost, but will change home: from separate assistants to skills that Gemini uses when needed. I do not know the detail of every step: the page says the conversion starts on 17 November and that until then you can keep using Gems.

1. Copy the instructions of your important Gems into a document of your own (for example in Google Docs).
2. Try skills right now, so there are no surprises in November: the Scopri di più (Learn more) and Crea skill (Create skill) buttons in the notice lead to the new feature.
3. After 17 November open the Manager and the Skills page and check everything works as before.

5. Skills

From the settings menu choose Competenze (Skills). The page explains it like this: you can create reusable custom instructions for more useful answers, Gemini automatically uses the relevant skills, but you can also apply them yourself by typing / in the box.

SCREENSHOT

Skills page with Create with Gemini, Create manually and recommended skills

The Skills page: Create with Gemini, Create manually, upload a skill, and the recommended ones.

• Crea con Gemini (Create with Gemini): describe in words what you want and Gemini writes the skill.
• Crea manualmente (Create manually): you write the instructions.
• The upload button: import a skill you already have.
• Recommended: ready examples, such as «Imitate your writing style» (it learns from texts you wrote in Workspace apps) and «Focus your energy».

5.1 How to write a good skill

A skill is like a recipe: it has a title, says when to use it, the steps and the result. For the pond you could write:

To call it by hand just type / and choose the name. Otherwise Gemini uses it on its own when the request looks like the case.

6. Which one to choose

The practical advice: write the instructions once, in a document of your own, and then paste them where needed. If one day Google changes the name of the containers, you already have everything in order.

Hands-on exercise

Thirty minutes to build, in order, the three levels.

1. Open Settings, then Personal context: read the two switches and write one or two fixed instructions with the Add button.
2. Open the Gem manager and read the 17 November 2026 notice. If you already have Gems, copy their instructions into a document.
3. Create a new Gem for a task you do often: write who, what, how, in what form. Save it only once and test it with Start chat.
4. Open Skills and choose Create with Gemini: describe a procedure you repeat often and let Gemini write the skill.
5. In a new chat type / and call the skill. Then ask the same thing without the /: notice whether Gemini uses it on its own.
6. Check the Gem manager to remove any duplicates.

Questions that come up in class

Do I lose my Gems on 17 November?

According to Google's notice, Gems start being turned into skills from that day and you can use them until they are converted. To stay calm, copy the important instructions into a document of your own.

What is the difference between a Gem and a skill?

A Gem is a separate assistant you call. A skill is a reusable instruction that Gemini can use on its own when needed, or that you call with the / bar.

Is memory compulsory?

No: it has a switch. You can turn it off and delete past chats from the Personal context page.

Can I share a Gem?

Yes, the editor has a Share button. If you attached files as knowledge, their titles will be visible, and you will be asked separately whether to share the contents.

How many instructions should I write?

Few and clear. Four or five well-written lines work better than a page of contradictory rules.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Gemini on the web: Gem manager, editor, Skills, Personal context

author's screenshots and tests, 7 October 2026

Google's notice in the Gem manager about the move from Gems to skills

visible on 7 October 2026, with the date 17 November 2026

1. The keys to some rooms

Until now Gemini knew only what you wrote. With connected apps it can look into your things, but only if you give it the keys.

Imagine a new assistant in your home. At first it does not know where things are and asks you everything. If you give it the key to the study, it finds the quote by itself. If you give it the key to the bedroom, it sees things it does not need. Connected apps work like this: each app is a room, and you decide which ones to open.

2. The Connected apps page

From the settings menu (the cog) choose App connesse (Connected apps). At the top the page says: «Connect your favourite apps for smarter help», and it has two tabs, Da Google (From Google) and Altro (Other).

SCREENSHOT

Connected apps page with Google Workspace, Gmail, Calendar, Docs, Drive, Keep and Tasks and other integrations

The Connected apps page: Google Workspace with six apps, and other integrations below. Interface in Italian.

The first card is Google Workspace, with a single switch and the description «Summarise, find and get answers from your content to support your project». Inside there are six apps, each with its own name to call it by.

Further down you find other integrations, for example YouTube Music and Gemini Notebook, each with its own switch. On my account, that day, YouTube Music was on and Gemini Notebook off: see how much it matters to check in person.

3. How to use them: the at sign

In the text of a request write @ followed by the app name. Gemini understands it must look there. Here are four examples designed for someone building a pond. I did not run them on my emails and calendar, so as not to show personal data: try them on your own account.

1. «@Gmail find the nursery's email with the quote for the EPDM liner and tell me the figure and the expiry date.»
2. «@Google Calendar tell me on which afternoons of the next two weekends I am free to dig.»
3. «@Google Drive summarise the file *Pond quote* in five lines.»
4. «@Google Tasks add the task *Order the bedding sand* for Friday.»

And remember the fourth step of the method from lesson 3: for important data (figures, deadlines) verify in the original app.

4. Privacy: what to connect and what not

With a good habit you risk little. Before turning a switch on, ask yourself these questions.

• Do I really need it? If you never use Keep, leave that key off.
• What is inside? A Gmail box full of customers' and colleagues' emails is a very delicate room.
• Work account? If you use Google Workspace, rules and permissions are decided by the administrator: ask first.
• Can I take the key back? Yes, at any time, with the switch. To see what has been done, check the Attività (Activity) page.

Diagram

Which rooms you open to Gemini

Which rooms you open to Gemini

5. Scheduled actions

So far Gemini has answered when you asked. With scheduled actions it works even when you are not there: you set a task and a time, and the result waits for you in the chat. It is like an alarm clock, but instead of a sound you find an answer.

SCREENSHOT

Scheduled actions manager with templates and the My actions section

The Scheduled actions manager: ready examples and the list of your actions.

From the settings menu choose Azioni programmate (Scheduled actions). The page explains that you can receive «daily news, creative inspiration or tools to develop your skills», ready in the chat at the scheduled time. Below there are ready-made templates, each with a + button.

To create your own click Nuova azione (New action) and describe what you want and when. In the Le mie azioni (My actions) section all the active ones will appear; on my account, that day, it said «You have not scheduled anything yet». An idea for someone with a pond: «Every Saturday morning remind me of the seasonal jobs for the pond and check the weather in my area.»

6. Gemini Live: talking instead of typing

The last piece is voice. Gemini Live is the mode in which you talk to Gemini like to a person, out loud, without typing. Google presents it mainly for phone apps: you can show Gemini the camera view or share your screen, and ask questions about what it sees.

A typical use, for someone with a garden: you point at a pond plant with spotted leaves and ask out loud «What is it, in your view?». The usual rules apply here too: it is a help, not a diagnosis. For a sick plant, compare with a nurseryman.

Hands-on exercise

Twenty minutes, with the greatest care over data.

1. Open Settings, then Connected apps. Look at which switches are on in your account and ask yourself whether you really need them.
2. Turn on only one app, the one that seems least delicate (for example Google Calendar or Google Keep).
3. Write a request with the at sign and ask something simple (for example: «@Google Calendar what commitments do I have tomorrow?»). Check in the app that the answer is right.
4. Open Scheduled actions and create a new action useful to you, with a time you can keep.
5. The next day read the result in the chat and decide whether to keep it.
6. At the end, switch off the apps you no longer need.

Questions that come up in class

Does Gemini read my emails even if I do not ask for anything?

With the switch on, it can use them when the request concerns them. If in doubt, switch the app off or do not enable it, and read the privacy notice for the Gemini apps and the Activity page.

Can I also connect apps that are not Google's?

The page has an Other tab for integrations different from Google's. What is there depends on your account and changes over time.

Are scheduled actions free?

They cost in terms of usage limits and their availability may depend on the plan. Check on your account.

What if Gemini gets an event or a task wrong?

It must be corrected in the original app. That is why it is best to have Gemini say what it will do and to check afterwards.

Does Gemini Live work on the computer?

Google describes it mainly for phones. To know if it is available on your computer, check the entries in the writing box and the help page.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Gemini on the web: Connected apps and Scheduled actions pages

author's screenshots, 7 October 2026

Google, guide to Gemini Live with camera and screen sharing

official description, not tried by the author for this course

1. An app or a website?

The website is like going to the bank: you open the browser, look for the address, go in. The app is the bank round the corner, always open.

Until yesterday, to use Gemini on your computer you had to open the browser and go to gemini.google.com. Since 10 September 2026 Google has made a Windows app available worldwide, for Windows 10 and 11. The advantage is convenience: an icon on the taskbar, a window of its own and, above all, a keyboard shortcut that opens it in an instant, wherever you are.

Important note: the app was already installed on the PC I use for this course, so I describe the installation part from Google's official page and not from a test of mine. Everything else I saw with my own eyes.

2. Installing it safely

The app is downloaded from the official page gemini.google/desktop. It is the only place to get it.

1. Open the page gemini.google/desktop by typing the address yourself in the browser, without clicking links received by email or found in ads.
2. Download the installer and run it. Windows may ask for confirmation: check that the publisher is Google.
3. At first launch sign in with your Google account, the same one you use on the web.
4. If you like, pin the icon to the taskbar: right-click the icon, then Pin to taskbar.

The program works with Windows 10 and 11. Some more advanced features, such as the Spark agent and the Omni video generator, according to Google require a Google AI subscription and an account of the right age.

3. The first tour: it is the same Gemini

Open the app from the taskbar. A window opens with the title Gemini, and inside you find something very familiar.

SCREENSHOT

Gemini app window with sidebar and writing box

The Gemini app for Windows window. Recent chats and personal data have been blurred. Interface in Italian.

The sidebar, the writing box, the model selector, the microphone: everything is the same as the website. The chat list is the same too: the conversations I had made from the browser, with the pond, the water lilies and the checklist, were already there, because chats are tied to the account and not to the program.

SCREENSHOT

Settings menu of the Gemini for Windows app with Personal context, Limits, Connected apps and other entries

The settings menu is also the same as on the web: Personal context, Usage limits, Connected apps, Gems, Gemini Notebook and so on.

• Same account, same chats: what you do in the app you find on the website and vice versa.
• Same settings: limits, connected apps, Gems, skills, Notebook.
• Same tools: the + button with attachments, images, video, music, Canvas and Deep Research.

4. Alt+Space: Gemini at the touch of a key

The real novelty of the app is the shortcut. According to Google, pressing Alt+Space from any screen opens Gemini over the program you are using, be it Word, a spreadsheet or the browser. According to the launch coverage, the key can be changed in the app's settings.

SCREENSHOT

Windows desktop with the Gemini window in the foreground

After Alt+Space on the desktop: Gemini appears in the foreground over the wallpaper.

On my PC, with the app already running, pressing Alt+Space from the desktop brought Gemini's window to the front, as you see in the screenshot. On someone else's computer, or with a different version, the panel may look more compact: follow what you see.

4.1 When it is handy

1. You are writing an email and want to polish a sentence: Alt+Space, paste, ask «make it friendlier».
2. You have a long text and need a five-line summary.
3. A quick question not worth searching on Google.
4. You need a list of ideas while in the middle of a job.

5. A conversation in the app

To try the app I did the same as on the website: I wrote a request and waited for the answer.

SCREENSHOT

Chat in the Gemini app with the answer on protecting a pond from cats and herons

The answer in the app: three solutions, each with «Come fare» (How to) and «Perché funziona» (Why it works).

After about fifteen seconds the answer was there, with three ideas, each split into «How to» and «Why it works»: the perimeter fence with invisible fishing line, the water deterrent with motion sensor and a third solution. As always, all to be verified: they are suggestions, not guarantees, and how herons and cats behave changes from garden to garden.

6. Web or app? The practical choice

Diagram

Web and app compared

Web and app compared

In practice you do not have to choose once and for all: you can have the app on your PC and use the website when you are elsewhere. The chats follow you everywhere.

Hands-on exercise

Fifteen minutes. If the app is not installed start from step 1; otherwise from step 3.

1. Type the address gemini.google/desktop in the browser and download the installer from the official page.
2. Install the app and sign in with your Google account.
3. Open the app and check that the chats you made on the website appear.
4. Open any program (for example a Word document), press Alt+Space and see what happens.
5. Ask a quick question from there and then close the window.
6. Pin the icon to the taskbar and decide for which tasks you will use the app and for which the website.

Questions that come up in class

Do I have to pay to use the app?

Not for the app itself, which uses the same account as the website. Some features (such as Spark and Omni) according to Google require a Google AI subscription.

Can I use the app on Mac or Linux?

The app I am talking about is for Windows 10 and 11. For other systems use the website.

What does Spark do?

Google presents it as a personal agent for tasks that need several steps, connected to apps such as Gmail and Drive. It requires a subscription: check on your account whether it is available.

Does the app read what is on my screen?

From the sources I consulted, Alt+Space opens Gemini over the active program; no automatic screen reading is mentioned. Always check the permissions the app asks for.

Can I change the Alt+Space shortcut?

According to the launch coverage yes, from the app's settings. If the entry is not there, your version may be different.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, The Gemini app is now available for Windows (blog)

launch of 10 September 2026; download from gemini.google/desktop

Gemini app for Windows running on the author's PC

screenshots of 7 October 2026

1. A new name, the same idea

If you knew NotebookLM, let me introduce its new name: Gemini Notebook.

On 16 July 2026 Google renamed NotebookLM to Gemini Notebook. The product remains a tool of its own, built for research, but the new name says it is more and more tied to the rest of Gemini. Old notebooks and old links, according to Google, are kept through automatic redirects: you do not have to redo anything. Browsing to notebooklm.google.com, in my browser I ended up on notebook.google.com.

Google says that more than 30 million people and more than 600,000 organisations already use it, for example for onboarding training materials or to turn notes into audio and video.

2. The Gemini Notebook home

Open notebook.google.com (or, from the Gemini app, the Nuovo notebook (New notebook) entry in the sidebar) and you enter the home.

SCREENSHOT

Gemini Notebook home with featured notebooks and a New notebook button

The home: featured notebooks at the top; recent notebooks (blurred here) further down. Interface in Italian.

• At the top the tabs Tutti (All), I miei notebook (My notebooks), Scopri (Discover) and Raccolte (Collections), the search, the switch between grid and list and your plan badge (here PRO).
• Notebook in primo piano (Featured notebooks): public examples, prepared by authors and organisations, which you can browse for inspiration.
• Notebook recenti (Recent notebooks): your own, with date and number of sources.
• The Nuovo notebook button to start.

3. Your first notebook

Click Nuovo notebook. A three-panel screen opens, which stays the same for every notebook.

SCREENSHOT

Untitled notebook with Sources, chat and Studio panels

An empty notebook: Sources on the left, chat in the middle, Studio on the right.

Diagram

The three panels of Gemini Notebook

The three panels of Gemini Notebook

3.1 Adding sources

SCREENSHOT

Window with Upload file, Websites, Books, Drive, Copied text and web search

The Add sources window: you upload files, websites, books, Drive documents or pasted text.

• Carica file (Upload file): PDFs, images, documents, audio and other formats.
• Siti web (Websites, also YouTube videos): you paste an address.
• Libri (Books): add a book from the linked library (the icon is the Google Play Books one).
• Drive: documents you already have in your Google Drive.
• Testo copiato (Copied text): you paste a text directly.
• Web search: you write what you are looking for and Gemini Notebook proposes sources, with the Web and Ricerca rapida (Quick search) options.

For this lesson I pasted as Copied text a small guide written by me on how to build a garden pond: a few lines, but with the three depth zones, the liner, the oxygenating plants and green water.

SCREENSHOT

Notebook titled Practical guide to building a garden pond with a summary

As soon as the source was inserted, Gemini Notebook gave the notebook a title, chose an icon and wrote a summary.

4. Asking the sources (with citations)

In the chat in the middle you can ask any question about your sources. I asked (in Italian): «At what depth should I put the water lilies and why?».

SCREENSHOT

Gemini Notebook answer on water lily depth with citation numbers

The answer with the citation numbers: clicking them shows the passage of the source the information comes from.

The answer was: water lilies go at 40-50 centimetres, in the intermediate zone, with the explanation of the three zones. Next to the sentences small numbers appear (the citations): clicking them takes you to the exact passage of the source.

The part I liked most is the honesty. My guide did not explain the biological reason for the depth for water lilies, and Gemini Notebook said so: «The guide does not specify further biological reasons for choosing this depth». It invented nothing: it stopped where the source ended.

• Under the answer you find Salva nella nota (Save to note): you keep good answers in a note.
• At the bottom there are suggested questions to continue.
• The page always reminds you: «Gemini Notebook can make mistakes, so check the answers».

5. The Studio: from sources to materials

The panel on the right, the Studio, turns your sources into other materials. Each entry is a button.

5.1 Example: the mind map

I clicked Mappa mentale (Mind map). A window opens where you choose the language, the sources to use and, if you like, a specific topic. At the bottom there is also the Utilizzo dell'AI (AI usage) meter (how much it uses) and two buttons: Genera più tardi (Generate later) and Genera ora (Generate now).

SCREENSHOT

Mind map window with language, sources, topic and AI usage

The window to create the mind map: language, sources, topic, AI usage.

SCREENSHOT

Pond mind map with four branches

The generated map, with four branches: Depth zones, Laying the liner, Water and plant management, Fauna management.

In about ten seconds the map was ready, with the notebook title in the centre and four branches. Clicking the arrow to the right of each node opens the sub-topics; there are buttons to enlarge, reduce and download the map, and two rating buttons («Quality content» and «Poor content»).

5.2 Example: the audio overview

SCREENSHOT

Window with the Deep dive, Summary, Critique, Debate formats, language and length

The Customise audio overview window: four formats, language, length and hints for the hosts.

• Deep dive (Approfondimento): an animated conversation between two hosts who analyse and connect the topics in your sources.
• Summary (Riepilogo): a concise overview to grasp the main ideas quickly.
• Critique (Critica): an expert review of your sources, with constructive feedback.
• Debate (Dibattito): two hosts comparing different perspectives.
• Then you choose the language, the length (Short or Default) and, if you like, what the hosts should focus on.

6. Gemini Notebook inside Gemini

The new name is no accident. Now Gemini Notebook and the Gemini app talk to each other.

• In Gemini's sidebar there is the Notebook section with Nuovo notebook, and in the settings menu the Gemini Notebook entry.
• In the Connected apps page there is the switch for Gemini Notebook (on my account, that day, it was off).
• According to Google, notebooks sync between the Gemini app and Gemini Notebook, and integration with Google Search (AI Mode) has been announced as the rollout widens.
• A novelty announced in July 2026: a cloud computer for every notebook, which runs code to analyse data and produce charts and spreadsheets. It was available immediately for Google AI Ultra and Workspace customers, with arrival on Pro plans expected in the following weeks. I did not try it: check whether it is already in your plan.

7. Ideas, limits and common sense

1. Study: you upload handouts and articles, generate a mind map and a quiz, and revise.
2. Work: you upload manuals or regulations and ask precise questions, with the citation telling you where to look.
3. Compare: you upload three quotes and ask for the differences, with a data table.
4. Learn on the move: the audio overview turns your notes into something to listen to in the car or in the garden.

Hands-on exercise

Twenty-five minutes, three or four sources on a topic you care about.

1. Open notebook.google.com (or click Nuovo notebook in Gemini's bar) and create a notebook.
2. Add three different sources on the same topic: a pasted text, a Drive document or a PDF, and a web page.
3. Read the automatic summary and compare it with what you already know: does it look right?
4. Ask two questions about the sources and click at least one citation to check the original passage.
5. Ask a question the sources do not answer and see how Gemini Notebook reacts.
6. Generate a mind map and save it. If you like, also open the audio overview window and try the four formats.

Questions that come up in class

Does NotebookLM no longer exist?

It changed name: since 16 July 2026 it is called Gemini Notebook. Old notebooks and old links, according to Google, stay reachable through automatic redirects.

Why use Gemini Notebook instead of the chat?

When you want answers based only on your documents, with citations. The normal chat also answers with what it knows, and sometimes invents.

How many documents can I upload?

It depends on your plan and changes over time: check the limits indicated by Google on your account.

Can I share a notebook?

Yes, there is a Share button at the top. The same cautions as Gemini's shared links apply: whoever receives it sees your sources.

Is the audio overview always correct?

No: it is a spoken summary produced by AI, which can simplify or get things wrong. Listen to it as an introduction and check the sources.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, NotebookLM is now Gemini Notebook (blog)

July 2026 announcement; usage figures and cloud computer

Google Workspace Updates, NotebookLM is now Gemini Notebook

July 2026

Gemini Notebook, notebook.google.com

author's screenshots and tests, 7 October 2026

1. From chat to agent

A chat gives you advice. An agent gives you the job done.

So far Gemini has written texts, created images, answered questions. But if it tells you «create a web page with the pond checklist», you still had to create the file yourself. Antigravity takes one step further: it is an agent, that is an assistant that can act on your computer. It reads the files in a folder, creates new ones, runs commands and reports what it did. Think of it as a site foreman: you say what you want, it organises and works, and asks permission before anything delicate.

SCREENSHOT

Antigravity window with New Conversation, Conversation History, Scheduled Tasks and writing box

Antigravity's start screen. Previous conversations in the sidebar have been blurred.

According to Google, Antigravity 2.0 is a standalone desktop application, available for macOS, Linux and Windows, built to work with agents and without the components of a code editor. Besides the app there are other pieces of the same world:

2. The model and the tank

Under the writing box there is the name of the model, for example Gemini 3.8 Flash High. Click it (or press Ctrl+/) and the list opens.

SCREENSHOT

List of models: Gemini 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.1 Pro, Claude Sonnet 4.6, Claude Opus 4.6, GPT-OSS 120B

Antigravity's list of models, on 7 October 2026.

On my account, that day, there were these models. Note that they are not only Google models.

The second word (High, Medium, Low) shows how hard the model tries to reason. The higher it is, the more accurate and the more it uses. For this course's tests I kept Flash High.

SCREENSHOT

Models & Usage page with plan, credits and limits

Settings, Models & Usage: the plan, the credits and the usage limits.

• Your Plan: my plan was Google AI Pro, with the Upgrade button to move to Ultra.
• Gemini Models: two limits, Five Hour (short) and Weekly, with the percentage left and when it refills. That day I had 98 per cent of the week.
• Claude and GPT models: they have separate limits.
• Model Credits: the *Enable AI Credit Overages* switch is used to spend paid credits when the quota runs out. It was off, and for a beginner it is better to leave it so.

3. Projects and environment: where the agent works

Before writing the task you must say where the agent can work. Above the box there is the No Project menu.

SCREENSHOT

No Project menu with New Project, Quick Start and No Project

The projects menu: New Project, Quick Start, No Project.

• No Project: no working folder. The agent works in a more limited way.
• New Project: you choose a folder on the computer, which becomes the project.
• Quick Start: a quick start to begin without configuring.

I chose New Project and pointed to a new, empty folder, created for the purpose and called *demo-antigravity*. The normal Windows window for choosing a folder opens.

SCREENSHOT

Antigravity with the demo-antigravity project selected

The selected project: the folder name appears at the top, the environment menu under the box.

Under the box there is also the Local menu (Ctrl+.). It is used to choose the environment.

SCREENSHOT

Local and New Worktree menu

The environment: Local or New Worktree (only for Git folders).

Local works directly in the folder. New Worktree creates a separate working copy, but is available only for Git repositories (the file history of programmers). For normal jobs Local is fine.

4. The first real job, step by step

Diagram

The work cycle with an agent

The work cycle with an agent

Let's do a simple, recognisable task. I wrote, in Italian:

SCREENSHOT

Antigravity writing box with the checklist request

The request written in the box, with the project selected.

I press Enter. Antigravity gives the conversation a title (here «Creazione Checklist Laghetto Giardino», Creating a Garden Pond Checklist) and starts.

4.1 Step 1: the permission request

After a few seconds the agent wants to look at what is in the folder. Before doing so it stops and asks.

SCREENSHOT

Allow listing directory contents request with five options

The permission request: «Allow listing directory contents?» with the command it wants to run.

The command here was Get-ChildItem -Path ., that is «list the files of this folder»: harmless. I chose 1.

4.2 Step 2: the work

SCREENSHOT

Antigravity showing reasoning and steps in progress

The agent reasons («Thinking for 12s») and shows the steps it is carrying out.

Antigravity shows what it is doing: «Ran Get-ChildItem», «Thinking...». You can open its reasoning. After about a minute the file was created: index.html, about 20 kilobytes, with almost six hundred lines.

SCREENSHOT

Antigravity with Edited index.html and a second permission request

The agent wrote the file («Edited index.html +592 -0») and rereads it to check; then it asks a second permission.

Note something I liked: the agent rereads its own work in pieces («Analyzed index.html #L1-100») and runs a check with Test-Path, asking permission again. I answered 1 again.

4.3 Step 3: the result

SCREENSHOT

Agent's final report with 1 file changed and Review button

The agent's final summary, with the Review button and «1 file changed +593 -0».

At the end the agent writes a report: what it did (checklist divided into six phases, progress bar, automatic saving of ticked boxes, quick buttons, green responsive style) and how many changes it made: 1 file changed, +593 -0, that is one file with 593 lines added and none removed.

RESULT

Web page Garden Pond Construction with progress bar and checklist

The page created, opened in the browser: title, progress bar and the first items with boxes to tick. Text in Italian.

5. Reviewing the work: the Review panel

The Review button opens on the right a panel with the changes line by line: in green what was added, in red what was removed. It is your safety net.

SCREENSHOT

Review panel with the code of index.html

The Review panel: the code of the index.html file with line numbers, all green because it is a new file.

1. Read the agent's report: does it say what it did?
2. Open the result (here, the page in the browser) and really try it.
3. Check numbers and facts. In my page, for example, the EPDM liner calculation comes to about 5.6 metres: redoing the sum with the depth you chose yourself, the result changes. An AI writes well, but sums must be redone.
4. Ask for corrections in the same conversation: «Add a section on costs», «Change the green to blue».

6. Commands with the slash and the at sign

In the writing box you can type / to see the commands, or @ to call something. They are shortcuts to tell the agent *how* to work.

SCREENSHOT

List of commands goal, schedule, browser, plan, grill-me, teamwork-preview, learn, boost

Typing / opens the list of commands.

The beginner's advice: for any task that is not trivial, start with /plan or /grill-me. You spend two minutes getting aligned and save half an hour of corrections.

SCREENSHOT

@ menu with Rules and Conversation

Typing @ lets you call the Rules or a previous Conversation.

With @ you call the Rules, your work rules, or a Conversation, that is a previous conversation to resume from. At the bottom right of the box there is also the Main Agent menu.

SCREENSHOT

Menu with Main Agent and two other agents

The Agents menu: the main agent and, on my computer, two specialised agents.

Main Agent is the basic agent. Other specialised agents may appear in the list (on my PC, two I already had installed): each is a collaborator with a precise skill, which the main agent can call.

7. Scheduled tasks

As in Gemini, in Antigravity too there are tasks that start by themselves. You find them in the sidebar with Scheduled Tasks (shortcut Ctrl+U).

SCREENSHOT

Scheduled Tasks page with no task configured

The Scheduled Tasks page: search, New button and, here, no task configured.

On my computer, that day, it said «No scheduled tasks configured». With New or with the /schedule command you create one, for example «every Monday morning check whether there are new checklists in the folder and summarise them». These tasks use up limits: few and precise.

8. Safety: the settings that matter

At the bottom left there is Settings. The first page, General, contains the rules of the game.

SCREENSHOT

Antigravity General settings with Security Preset and Plan Review Policy

Settings, General: execution, global permissions, agent behaviour and network permissions. Account name and email are blurred.

Another useful page is Customizations: it shows your skills, subagents and MCP tools, with a token budget (here 20,000). If you exceed it, Antigravity automatically excludes the larger customisations.

SCREENSHOT

Customizations page with token budget and skills

Customizations: the token budget and the list of skills. A warning says the budget has been exceeded.

9. What it is really for

1. Pages and small tools: an interactive checklist, a calculator, a page for your shop.
2. Tidying and transforming files: reorganising a folder of documents, merging tables, renaming files in bulk.
3. Checking and summarising: reading a folder of documents and preparing a summary.
4. Learning to program: asking the agent to explain line by line what it wrote.

It is not the right tool for everything: to write a text or a piece of research, chat or Canvas are enough and use less. Antigravity is for when there is a job to do on the computer.

Hands-on exercise

Thirty minutes. You need the Antigravity app installed and an empty folder created for the purpose.

1. Create on your PC a new, empty folder, for example *antigravity-test*, away from your documents.
2. Open Antigravity, click No Project, choose New Project and point to that folder.
3. Write a simple task, for example: «Create an index.html page with a checklist for my hobby, with boxes to tick».
4. When the agent asks for a permission, read the command and answer 1, Yes allow this time.
5. Open Review, look at the added lines and then open index.html in the browser.
6. Ask for a change in the same conversation (a colour, an extra section) and check again.
7. Open Settings, then Models, and look at your plan and limits: how much have you used?

Questions that come up in class

Do I need to know how to program to use Antigravity?

Not for simple jobs: describe what you want in plain language. But you must be able to check the result, and that takes a bit of curiosity and a lot of caution.

What is the difference between the app and the IDE?

The Antigravity 2.0 app is built to work with agents, without a code editor. The IDE is a separate code editor, which you install from the Install IDE button.

Can I use models other than Gemini?

In my list there were also Anthropic models (Claude) and an open OpenAI model. The list depends on plan and period.

Can the agent delete my files?

If you give it permission, in theory yes. That is why you work in a dedicated folder, with backups, answering «allow this time» and reading every command.

Is it free?

It works within the limits of your Google AI plan (Five Hour and Weekly) and, if you turn it on, with paid credits. Check the Models & Usage page.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, Introducing Google Antigravity 2.0

May 2026 announcement (desktop app, IDE, CLI, SDK)

Antigravity app on the author's PC: models, settings, test project

screenshots and tests of 7 October 2026

1. Gemini or Gemma?

Gemini is a restaurant: you order, they serve you, you pay (or not). Gemma is the recipe: you take it home and cook yourself.

So far we have talked about Gemini, which lives in Google's computers: you write, the request travels over the internet, the answer comes back. Gemma is another family of Google models, with a fundamental difference: it is open-weight. The «weights» are the numerical heart of the model, and when they are open you can download them and run them on your own computer.

Diagram

Gemini and Gemma compared

Gemini and Gemma compared

2. Why run it locally

Running a model on your own computer is called locally. It has real advantages and equally real limits.

3. The features of the Gemma families

Here is a taste of what the family offers, without technical details. The data come from Google's and Ollama's official pages consulted on 7 October 2026.

3.1 Gemma 4 (April 2026)

• Released on 2 April 2026, under the Apache 2.0 licence (very permissive, also for commercial use).
• Available in several sizes: E2B and E4B (the smallest), 26B mixture-of-experts and 31B dense. In June 2026 a 12-billion-parameter version was added.
• It is multimodal: it also understands images and video, and the small sizes also audio.
• Context up to 256 thousand tokens in the larger sizes (the «memory» of a conversation).

3.2 Gemma 3 (the previous generation)

• Very small sizes: 270 million, 1 billion, 4, 12 and 27 billion parameters.
• Supports over 140 languages.
• From the 4-billion size up it also understands images.
• The smallest versions are meant exactly for computers with few resources: 815 MB the 1-billion model, 3.3 GB the 4-billion one.

Diagram

The sizes of Gemma 3

The sizes of Gemma 3

4. How it is launched, in broad terms

The simplest way to run an open model is a free program called Ollama. It is a bit like a music player: you install the player once, then you choose the tracks, that is the models.

• You download Ollama from the official site and install it.
• You choose a model from its online library, looking at size and requirements.
• You launch it with a command, for example ollama run followed by the model's name, and a chat opens in the terminal or in the app window.

I have Ollama on the computer I use for this course, but with code models, not with Gemma, and I did not try Gemma for the lesson: here I only tell you what is there and what to expect. If you want to try, start from the smallest model you find, and see how it answers on your computer.

5. When it makes sense (and when not)

For most people, Gemini in the cloud remains the most convenient and most capable choice. Gemma is fascinating for those with privacy needs, for those who study and for those who want to build something of their own.

Hands-on exercise

Fifteen minutes, without downloading anything: just read and decide.

1. Open the official Gemma page and the Ollama library page and read the available sizes with their dimensions.
2. Check how much memory and which graphics card your computer has (in Windows: Task Manager, Performance tab).
3. Write three things you would gladly do locally and three you prefer to do with Gemini in the cloud.
4. Decide: do I try Gemma, yes or no? If yes, which size and when?
5. If you decide to try, download only the smallest model first, and only from official sites.

Questions that come up in class

Is Gemma free?

Yes, Gemma models can be downloaded and used, and the Gemma 4 licence is Apache 2.0. The real cost is your hardware and electricity.

Is Gemma the same as Gemini?

No. They are different Google model families. Gemma is open and downloadable; Gemini is closed and used from the cloud.

Do I need a powerful graphics card?

For small models less is enough, but in general the bigger the model, the more memory and speed you need. Start from the small sizes.

Does it work without internet?

Yes, after downloading the model. To download it you need a connection.

Can I connect it to Gmail or Drive?

Not directly like Gemini. Locally you have the model, not Google's connected tools.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, Gemma 4: Byte for byte, the most capable open models (blog)

announcement of 2 April 2026

Google AI for Developers, Run Gemma with Ollama

official page

Ollama, library of the gemma3 and gemma4 models

pages consulted on 7 October 2026 for sizes and file dimensions

1. The path in brief

You made it to the end. Let's retrace the road: from «what are all these names?» to an agent building a page in your folder.

The thread connecting everything is simple. First you learned to choose (models, plans, limits) and to ask (four steps). Then you used the tools to search, study and create. Then you personalised Gemini and connected it to your things. Finally you brought it into the computer, with the app, with the document notebook and with an agent, and you saw that a local route also exists.

Diagram

The path taken, in four stages

The path taken, in four stages

2. The conclusions: five ideas that remain

1. Choose the right tool for the job. Do not send the truck to buy bread: Flash for everyday things, Pro for difficult tasks, Deep Research for research, Notebook for your documents (lessons 1, 4 and 9).
2. Write well and always verify. Context, goal, format, verification. Gemini writes sentences that sound right: checking a fact on a second source remains your job (lessons 2 and 3).
3. Give the keys with judgement. Every connected app and every permission to an agent is an open room: few, for the time needed, and always «allow this time» at the beginning (lessons 7 and 10).
4. Write instructions once, and keep them with you. Context, Gems and skills change name: a copy in a document of yours keeps you safe (lesson 6).
5. Start small and really try it. Free plan, small size, empty folder: you learn by doing, and you grow only when needed (all lessons).

3. The most common mistakes to avoid

I collected them from the frequently asked questions of the lessons, from the most serious to the most harmless.

4. And now? An action plan

4.1 The first week

• Redo one exercise of your choice, with your real project.
• Write in a document of yours your instructions for Gemini (context, a Gem, a skill).
• Open the Usage limits page and the Connected apps page and check that everything is as you want.

4.2 The first month

• Complete the mini-project below.
• After 17 November 2026, reopen the Gem manager and the Skills page and check your instructions are fine.
• Try Gemini Notebook on a real study or work topic.
• If you are interested, take a first tour with Antigravity on a dedicated folder.

4.3 To keep learning

• Google's official pages on Gemini, Gemini Notebook, Antigravity and Gemma: news arrives there first.
• The gianlu.it website, where you find the other handouts and contacts.
• The course when something changes: if a menu is gone, look for the feature by name and follow the criterion.

The final mini-project

One hour, your fixed project and four tools from the course.

1. Notebook. Create a notebook in Gemini Notebook and add three sources on your project. Generate a mind map.
2. Deep Research. Ask a question with context, goal and format. Open at least two sources and check one fact.
3. Gem or skill. Write the instructions (who, what, how, in what form) for an assistant or a procedure of your project, and also save them in a document of yours.
4. Canvas. Create a summary document of the project, edit it and export it to Docs.
5. Optional: Antigravity. In a new, empty folder, have it create a page or a checklist for your project, with «allow this time» and Review.
6. Self-assessment. Tick the boxes below.

Questions that come up in class

Where do I restart from if everything has changed after a few months?

From the four steps (context, goal, format, verification) and from lesson 1: if the difference between model, app and plan is clear to you, you find your way even with new names.

What is the most important thing in the whole course?

Verifying. Gemini saves you time writing and searching, but deciding and checking remain your job.

Should I move to a paid plan?

Only if you often hit the limits or need specific features, such as video. Look at the Usage limits page before deciding.

Will the screenshots stay valid?

Only in part: Gemini changes fast. Each lesson gives the verification date; for data that age (plans, prices, models) always check Google's official pages.

Where can I ask for help if I get stuck?

Restart from the lesson and its exercise; then write on the gianlu.it website, where you can leave a comment or contact me.

Lesson glossary

The words you meet in this lesson, explained in one line.

Further reading

The sources used to build this handout: if a figure interests you, go to the source and check it.

Google, official pages on Gemini, Google AI plans, the Windows app, Gemini Notebook and Antigravity

consulted on 7 October 2026

Google, Gemma 4: Byte for byte, the most capable open models (blog)

announcement of 2 April 2026

The author's tests and screenshots on Gemini, Gemini Notebook, the Windows app and Antigravity

7 October 2026