
Turn a small set of meeting notes into a useful result. In this beginner lesson, summarize the notes in Chat, ask Work to save a Markdown checklist, open the actual file, and check that its actions, constraints, and unknowns match the source.
In this lesson
Practice: Repeat with a shopping checklist.
Choose the correct Linux package, inspect the installation, launch the signed-in app, and verify one small local-file change.
In this lesson
Practice: Find the install instructions for your own operating system.
Enter requests, inspect Pages and local files, and return to existing work through panels, Recents, windows, and keyboard entry points.
In this lesson
Practice: Open a second chat and return to the first.
Attach source notes, write a clear request, answer a clarifying question, and check the saved brief against its source before a focused refinement.
In this lesson
Practice: Ask for the same result as a checklist.
Compare a vague request with a precise one, steer a running task with a correction, queue a follow-up, and revise a finished brief into a separate file without losing facts.
In this lesson
Practice: Change the audience and preserve the same facts.
Open the model and reasoning-effort controls, run one small task at High and at Medium with the identical prompt, read the context indicator, see the Speed setting, check usage limits, and restore GPT-6.1 Sol at High.
In this lesson
Practice: Choose a lower-cost available setting for a simple rewrite.
Attach a notes file and a screenshot, ask about one region of the image with the input named for each answer, then name a local file path and a public link in one request and check every source label in the saved result.
In this lesson
Practice: Replace one input and ask for the changed conclusion.
Create a project from two local folders, choose the primary folder on purpose, trust only known folders, and hand work from one chat to the next through a saved file. Then compare where a chat outside any project saves its note.
In this lesson
Practice: Start an unrelated chat without the project.
Create a sidebar section, try sorting and grouping, rename a chat and mark it unread, look at its share and copy options, archive it, restore it from Settings, pin it, find a phrase inside it and search for it by name.
In this lesson
Practice: Find yesterday's demo chat by a phrase from its contents.
Open a small web page as source and as a live preview, pin one comment to its heading, apply it, compare the change with an untouched copy and accept it. Then fix an undefined command in a tiny LaTeX file and download the real PDF.
In this lesson
Practice: Annotate one label instead of rewriting the whole file.
Create and trust a local project, ask a task where it runs and what it can access, save a file on this computer, run the same notes in a cloud task with an explicit attachment, and test that a local file is not visible in the cloud. Then check what Settings offers for sync on this build.
In this lesson
Practice: Explain why an unuploaded local file is unavailable in a cloud-only task.
Compare the three permission modes and keep Ask for approval, stop a run that has just started, send a narrower request, read a real terminal approval card and allow its one read-only command once, then see what Full access allows before confirming it.
In this lesson
Practice: Request a new output in the demo output folder with the required permission.
Ask for a small comparison built only from official websites, with a link and a page date for every source. Open the sources yourself, tell page dates apart from release dates, then check a short draft against its primary source and correct only the false claim.
In this lesson
Practice: Compare Linux Mint and Ubuntu using the same official-source criteria.
Open a tab beside the chat, mention Browser so the agent can find an official instruction page and return a sourced checklist, annotate one paragraph and ask about it, then check browsing history, downloads, site permissions and the pane layout.
In this lesson
Practice: Navigate to the official documentation for changing the desktop wallpaper and verify one instruction.
Install the ChatGPT extension in Chrome after reading its permissions, pin it, ask the side chat to summarize an open documentation tab, and let it follow a link in that same tab. Then check the Chrome permissions in the desktop app and mention an open Chrome tab in a new chat.
In this lesson
Practice: Repeat with another public documentation tab in the same connected browser.
Check which apps Computer use can control on this Linux desktop, send one complete request to rewrite a short Google Doc in Chrome, watch the agent type and correct its own first pass, then verify both the document and the matching text file it saved.
In this lesson
Practice: Change one checklist item in the same browser document and verify the saved result.
Learn what Appshots are and where they are available, then use the route that works on Linux: paste an ordinary screenshot of one window into a new chat, preview it, ask one specific question, and compare the answer with the real window.
In this lesson
Practice: Share a different safe window and compare the context.
Find the voice controls in Settings and in the message box, learn the difference between dictation and a live voice chat, choose a voice, and see what really happens when voice chat starts on a computer without a microphone. Then complete and revise a small checklist by typing.
In this lesson
Practice: Find the dictation control and choose a voice setting, then request one checklist revision.
Give one request with exact paths: copy three notes into a new folder, keep the originals, and build an index with one line per note. Read the index in source view, rename it, confirm each copy matches its original, and open an original to check it.
In this lesson
Practice: Change the output naming convention.
Open the integrated terminal and look around a folder with cd, pwd and ls. Ask the agent to write a tiny reusable Python script that totals a CSV column and to explain the command before running it. Then run the same script yourself on two files and read both saved results.
In this lesson
Practice: Run the same script with one changed expense.
Ask the agent to open a short PDF plan and extract its action rows without guessing, so the undecided date stays undecided. Then create a one-page checklist PDF from those rows, compare it with the source, and make a revised copy with a larger heading.
In this lesson
Practice: Request a larger heading and inspect the revised PDF.
Turn short meeting notes into a one-page Word document with clear headings, a small action table and the unknowns kept as unknown. Read it at full size, then rewrite only the introduction for a new team member in a separate file, and compare the two versions.
In this lesson
Practice: Change the audience to a new team member.
Build an Excel workbook from a five-row CSV file: the original rows, a blank category flagged as Unknown, a SUM total, SUMIF totals by category and a linked column chart. Click the totals to read their formulas, then add one row in a separate copy and check that the formulas and the chart follow.
In this lesson
Practice: Add one expense and recalculate.
Ask for exactly three editable slides built only from meeting notes, in a plain readable style, with the unknowns kept. Click through each slide and its speaker notes. Then change only slide 2 to a two-column layout in a separate copy, select it yourself when the agent cannot confirm it, and check that the other slides did not change.
In this lesson
Practice: Change only the second slide's layout.
Describe one simple object clearly and save the generated image as a PNG. Leave a comment on the image to use it as a reference and change only the color, compare both versions in Canvas view, then ask for a transparent-background cutout saved as its own file and check it.
In this lesson
Practice: Change one color while preserving the composition.
Ask Visualize for a small workshop-cost calculator with two number inputs and a clear formula. Check the starting total, change each input and confirm the math yourself, then add a travel-fee input and test it the same way.
In this lesson
Practice: Add a separate fixed travel fee and test the result.
Ask Sites for a tiny one-page resource list with three links and a checklist, and check its preview before anything goes live. Change one link label with an annotation, save it as version 2 and deploy it privately. Then look at the saved versions, see that checklist ticks are not saved, and change the site's short address name.
In this lesson
Practice: Revise one link label and compare saved versions.
Ask Pages to save a small checklist Page and a second Page with two questions. Group both under a private parent Page, then find the parent again in the Pages list, without the chat, and open each Page from it.
In this lesson
Practice: Create a second small Page and find both.
Add a heading, a table, a link and a reusable prompt to a saved Page. See that clicking the prompt sends only its placeholder, then use it on four real notes and save the result on a linked child Page. Add a file reference, move a link, write one line with Generate, and open the file and the child Page from the checklist.
In this lesson
Practice: Move the checklist above the table.
Change only the Unknowns section of a Page and check that the rest stays the same. Look at the toolbar for selected text. Ask the agent to add a comment, reply to it and resolve it, and see that resolving does not answer the question. Then check who can see the Pages, without changing anything.
In this lesson
Practice: Ask for a shorter selected paragraph rather than replacing the Page.
Find the Gmail plugin under Customize, install it and connect it to your account. Read what its tools can write and what they can only read, and use its on and off switch. Then use it with @Gmail to read one email without changing anything, and look at its four permission choices.
In this lesson
Practice: Invoke the same plugin explicitly in a new chat.
With a Google account connected, ask for one precise change at a time: one heading in a Google Doc, two cells in a Google Sheet and one slide title. Check each result in its own service. Then read an email and a calendar event, look at a draft that is not saved, and create the event only when you ask for it.
In this lesson
Practice: Repeat the same scoped pattern with a different allowed app.
Look at the skills installed on your computer, read one skill's page, switch it off and on, and open its folder. Then install the PDF skill with Skill Installer, call it with a dollar sign, and check the one-page PDF it makes from meeting notes.
In this lesson
Practice: Run the same skill on a second short note.
Make one small, repeated task reusable in three ways. Create a skill with Skill Creator, test it, and add a Due column. Create a document template from a reference document and use it for new notes. Then prepare a plugin package in Work on the web, and read what is installed and what is only prepared.
In this lesson
Practice: Change the output format and test again.
Do a small task once: turn meeting notes into a checklist, and check it. Then describe that workflow to Skill Creator, with input and output paths you choose and a check before saving. Run the saved skill on different notes, and compare a result with its source, fact by fact.
In this lesson
Practice: Refine the generated skill's naming rule.
See three ways the agent adapts to you. Give a preference in one chat, find the memory settings and the switches for a single chat, give a file as context and ask where the details came from, and save a personal instruction that a new chat follows with memory off. Then look at the same kind of settings in ChatGPT on the web.
In this lesson
Practice: Disable memory use for a demo chat and compare the behavior.
Make a project from a small code folder and trust it only because you know what is in it. Ask for a beginner's guide without changing or running anything. Then look for yourself: open the page file and the script, open the page in a browser and click its button, and ask a terminal where you are and whether anything changed.
In this lesson
Practice: Ask which file would change the page heading.
Ask the agent to explain a small page without changing it. Turn on Plan mode, read a three-step plan and skip implementing it, ask a quick question in a side chat, save the agreed requirements in a file, and compact the chat. Then check that the agent still knows the details.
In this lesson
Practice: Change one requirement before implementation.
Describe a small web page in plain words, open it in the built-in browser and click its button. Then mark the button with an annotation, ask for one color change and nothing else, reload, click again, and read the two lines that changed.
In this lesson
Practice: Change only the button label.
A small script adds up five expenses and prints the wrong total. Show the agent what you ran, what it printed and what you expected, ask for the smallest fix with an automated check, read the change before you accept it, and run the script and the test again yourself.
In this lesson
Practice: Change one amount and check the result again.
Ask the agent for two small changes to a page, then review them yourself. Read the diff hunk by hunk, leave a comment on the line you do not want, stage the change you want, revert the other one, and compare the branch with main. Nothing is committed in this lesson.
In this lesson
Practice: Leave an inline comment asking for one correction.
Make one small change on a new branch and read the diff. Ask the agent to commit it, push the branch and open a draft pull request. Then review it: mark the file as viewed, ask what changed and which checks ran, post a comment, save review instructions, and find the pull request in the Code Review list.
In this lesson
Practice: Find your demo PR in the review inbox and reopen its changed file.
Check Worktree to give one chat its own copy of the project, and change one line there. Compare it with main and see how a local settings file is copied. Run a second change in a second worktree, compare all three folders, and delete a worktree in Settings.
In this lesson
Practice: Create a second independent change in another worktree.
Run two small checks in two separate chats. Then ask one chat to use two subagents in parallel, one per file, and read the combined summary. Open the Subagents panel to see each subagent's own answer, and send one of them a follow-up.
In this lesson
Practice: Ask one agent to recheck a single finding.
Make one small change in a worktree chat. Fork the chat into a new worktree and check that the fork remembers the request. Then move the fork to the main checkout and check the folder, branch and file in the terminal. Last, see where else a new chat can run.
In this lesson
Practice: Fork the chat for an alternative tiny heading change.
Write project instructions in AGENTS.md and see a new chat follow them. Save a setup script and a one-click Check action. Add a rule that forbids git push and test it. Then review a project hook, trust it, see it record one event, and switch it off.
In this lesson
Practice: Change one project instruction and verify it in a new chat.
Create a cloud environment for a small practice repository and look at its settings. Add an environment variable, prepare a setup with a check, publish it, and run a fresh cloud task. Then publish a second revision and compare a fresh task with an existing one.
In this lesson
Practice: Start a fresh task from the same environment.
Find an SSH connection in Settings and create a project from a folder on that computer. Answer the trust question, run a small task on the remote computer, see a failed try and a retry that passed, and open the remote file. Then disconnect, see the file become unavailable, and connect again.
In this lesson
Practice: Reconnect and revise the same remote file.
Test a request by hand, then save it as a scheduled task with a time zone, a new chat for each run, and a model. Run it now, stop a run that went wrong, change the task in plain words, pause and resume it, run it again, and delete it.
In this lesson
Practice: Change the topic to space technology and confirm the saved update.
Start a slow five-minute demo, then ask the same chat to check back every minute, speak up only when it matters, and stop when it is done. Read the follow-up card settings, see a quiet check, the result, and the follow-up pausing itself.
In this lesson
Practice: Change the notification condition and inspect the saved task.
Type /goal to give the agent an objective it keeps working on: three summaries and an index, planned first and then created. Pause the goal, add a requirement, save and resume it, and check the five files it saved. Then give a small goal that waits for your approval before it changes anything.
In this lesson
Practice: Pause the goal and change one output requirement before resuming.
In Work on the web, create tasks that start when something happens in a connected service. An email trigger with an exact subject filter never ran and was paused; a GitHub comment trigger fired and reported the comment time; then both tasks were cancelled.
In this lesson
Practice: Tighten the sender/channel/repository filter.
A task that cannot finish says what it needs. Here a request named a file that did not exist; the answer asked for the file, the same chat created it, and the next answer matched its three lines. Then the Activity view, notification and sleep settings, Help, feedback and /status, and archiving a chat and restoring it.
In this lesson
Practice: Disable noisy demo notifications and confirm the setting.
Change how the app looks and set it back, choose between steering and queueing follow-ups, set a pop-out hotkey and clear it, find every shortcut, show a pet in the Mini and hide it, and read your token activity.
In this lesson
Practice: Set one useful shortcut and restore its default.
Create a dot, an always-on agent with its own cloud computer, choose where it can work, give it a name and a look, and hand it one small job with a source, a scope and a clear stop. A first request arrived cut short; a shorter retry and a second job came back with sourced answers.
In this lesson
Practice: Change the focus to space technology in the same conversation.
Your dot works on its own cloud computer, with its own browser. You can open that computer, take over and give control back. Separately, you can allow your own computer and connect apps like Google Drive, and take that access back. Offline is a different state from revoked.
In this lesson
Practice: Revoke the demo local grant and inspect the resulting limitation.
Message your dot in ChatGPT and through a contact method like Slack. A call did not connect, so the correction was typed instead. Slack was connected, one item was sent there and checked, and a newer dot returned a short file to open and download.
In this lesson
Practice: Ask the same dot for a status update through another available channel.
Your dot can hand work to separate tasks, each with its own chat and report file. Track them in the Activity view, correct one inside that task, run one in a prepared cloud environment, and have the dot continue an exact task from its deeplink.
In this lesson
Practice: Change the scope of just one delegated task.
Your dot can work on a schedule or wait for an event. A recurring check was saved on a retry, delivered its first result by itself and was canceled. Its card showed the schedule and its end. A Gmail monitor never ran; a Slack monitor ran once and paused itself. A short practice stopped when a check was blocked.
In this lesson
Practice: Change the topic and reporting threshold, then confirm the saved update.
Add a custom rule so your dot asks before a specific action, review and approve that action once, pause and resume the dot, stop a running task, pause a schedule, and remove the rule and access you no longer need. A first resume failed and a second worked.
In this lesson
Practice: Change one rule and verify its effect with a harmless draft request.
This course contains the use of artificial intelligence.
Have you only used ChatGPT to type a question and read a reply? This course shows you, step by step, how to get real work done with the ChatGPT desktop app, with no coding, terminal, or automation experience needed.
Every one of the 60 short lessons is hands-on. You watch real controls in the real app, see an actual result, and finish with a small practice task you can repeat on your own. New terms are explained while you point at the real thing, and there are no large capstone projects.
What you will learn
Start from zero: install and launch the app, find your way around, and complete your first real task in Work.
Context, projects, files, and permissions: attach the right material, keep related chats in a project, and control what the agent may access and do.
Search, browsers, Computer Use, and voice: research with sources you can check, let the agent work in a visible browser, and understand what each tool can and cannot do.
Everyday files: read, create, and revise PDFs, Word documents, spreadsheets with formulas and charts, and three-slide presentations.
Images, visualizations, sites, and Pages: generate and edit images, build a small calculator, publish a private one-page site, and organize Pages in Space.
Plugins, connected apps, skills, and memory: connect Gmail, Docs, Sheets, and Calendar safely, install and create reusable skills, and control personalization.
Codex for complete beginners: understand, change, test, and review a tiny code example, then use Git, worktrees, subagents, and a draft pull request.
Parallel work, remote access, and schedules: run work in the cloud, over SSH, on a schedule, or until a goal is done, and recover a stuck task.
Your Dot: assign bounded, ongoing work to an always-on agent, and control it with rules, approvals, and stop controls.
How the course works
Each lesson has real footage of the app, English narration and captions, a short quiz to check what you remember, and a repeatable practice task. After every section, a short AI role play lets you practice explaining and deciding in your own words.
You only need a ChatGPT account with access to the desktop app. The lessons were recorded on Ubuntu Linux, and the lessons point out where controls differ between systems and builds. Features and menus can change between app versions, so each lesson shows how to find the control rather than relying on memorizing a screen.