Can a ChatGPT Dot help teachers and students? Lesson decks, course follow-ups and tutoring calls: 27 real cases

The short version: where a Dot helps teachers and learners, how to set it up, where it gets stuck

Is a Dot worth it for teaching and study?

Yes for the work around a course; no as a child's own assistant, and no as a stand-in for the learning itself. In the cases below it helps in three ways. First, it keeps track: a course portal read twice a day, Google Classroom notices and Slack announcements turned into one list of what is due and what is still open, and trips that clash with a class (cases 7, 9, 24 and 28). Second, it prepares first drafts that a teacher checks: a first review of a student's project (case 1), a 60-page lecture deck built from the teacher's own course site (case 2), scripts for 40 lecture videos (case 17), a revised class notice (case 19), and post-course replies with a one-page handout (case 26). Third, it practises with you on a call: English drills, a chess game talked through move by move, and test revision with a parent next to the child (cases 5, 20, 21 and 22). It gets stuck at school sign-ins and final confirmation pages, when nobody wrote down what isn't decided yet or what a good result looks like, when it isn't told that a task is finished, and in live practice with children when no adult is steering. It suits teachers, trainers, adult learners and parents whose classes already sit in a calendar and whose course mail arrives in Gmail, and who still mark, send and hand in work themselves.

How to set it up

1. Use it as an adult. Dots are not yet available to users under 18, so the Dot belongs to the teacher, the adult learner or the parent (Dots privacy FAQ). Schools on ChatGPT Edu can try the dots beta once a workspace admin turns it on (Introducing dots), and in Business, Enterprise and Edu workspaces data is not used to train models by default.
2. Connect mail and calendar first. Most course news arrives by email, including Google Classroom notices (case 28), and the calendar doubles as the timetable for clash checks (cases 8 and 9). Connect Gmail and Google Calendar, then add Slack or Notion if your classes announce things there.
3. Sign it in to the course portal once. When a school site has no connector, the Dot sends you a private sign-in form or lets you take over its browser (Computers and apps). simon_vaerion_roc and Teto66 did this for their university systems (cases 7 and 28). Check one read by hand before you schedule a twice-daily check.
4. Point it at your own material: the class slot in your calendar, the course site, slides, a repo, a client survey, or a scanned PDF of the textbook pages (cases 1, 2, 15 and 20). yenshilun58's Dot built his deck from his own course site and cross-checked it, and he found the content accurate (case 2).
5. Set Custom rules so that sending email, booking, publishing and submitting stay on Ask before taking action (Getting started with your dot). Eve Cogan's class was booked by email before she had read the message (case 13); Snapper AI's rules make the Dot ask first (case 11).
6. Write down what is undecided, what must stay the same, and what a good result must include. Add "don't make up anything that isn't decided" and "don't send anything" (cases 19 and 26), plus the one thing that has to happen, such as "everyone types into an AI at least once" (case 18).
7. Tell it when you finish something, and read the first result before you rely on it. Teto66 tells her Dot when an assignment is submitted, so it marks the task done and stops the reminder (case 28). Antonio Leiva still writes the feedback himself after the Dot's first review (case 1).

Where it gets stuck, and what worked

1. A school site with no connector. What worked: sign its cloud browser in through the private sign-in form (cases 7 and 28), or let it install the tool it needs, as Antonio's Dot did with Whisper to transcribe a student's video (case 1).
2. The last confirmation page. Teto66's Dot filled a city library reservation basket but couldn't open the final confirmation page because of a browser security limit, so she confirmed on her phone; at the university library its renewals went through, and it said which books couldn't be extended (case 28). What worked: let it take the job to the last step and finish that step yourself.
3. Reminders for work you already handed in. What worked: tell it when you submit. Teto66's Dot then marked the assignment done, stopped the reminder, and still flagged a new attendance report as a separate task (case 28).
4. Undecided dates and details filled in. What worked: say what isn't decided. GAKU's deadline stayed "undecided", and none of Akira Kudo's draft replies carried an invented date or link (cases 19 and 26).
5. A plan that meets every condition and misses the point. Kohei Uesugi's workshop plan needed no AI on the day, though that was what he wanted (case 18). What worked: write down the one outcome that must happen.
6. Sending before you check. Eve Cogan's booking email went out unreviewed (case 13). What worked: Custom rules that ask first, plus "don't send anything" in the request (case 26).
7. Live practice drifting. Without rabbiekao steering, the children in her tutoring class drifted into asking for jokes and songs (case 5). What worked: an adult steers in the room; adult learners ask why each move made sense and keep a transcript to review (cases 21 and 22).
8. Uploads and slow turns. Three scanned pages at once gave a file-size error, and one or two at a time worked (case 27). Each step involves some waiting, so it doesn't suit questions that need an answer right away (case 26). What worked: hand over the job before bed, as yenshilun58 and the parent in case 27 did, and send pages in small batches.

What it can do now, and what it can't yet

It can now:

• Read a course portal, Google Classroom notices in Gmail and Slack announcements, and tell you what's new, what's due and what's still open.
• Check trips and events against your timetable, and point out when a class notice and your calendar disagree on a room.
• Build a first draft of a lecture deck, lecture scripts, a session plan or a handout from your own material.
• Give a first review of a student's project, including transcribing a video.
• Sort post-course messages into a table with draft replies, without inventing dates.
• Renew library books on a site you signed it in to, and report the new due dates.
• Practise a language or a game with you on a call, make a review sheet overnight, or build a practice site on ChatGPT Sites.

It can't yet:

• Be a child's own assistant; Dots are not available to users under 18.
• Know you've handed something in unless you tell it.
• Get past every final confirmation page, sign-in or code on its own.
• Keep live practice with children on track without an adult steering.

One message to start with

Be my assistant for this term. Every morning and evening, check my [university Gmail], [Slack channel] and [course portal; I'll sign in when you send the form] for new assignments, class changes and deadlines. Keep one list of what's open and what's done, and remind me [2] days before each deadline. When I tell you I've submitted something, mark it done and stop that reminder. If a notice and my calendar disagree on a time or room, tell me. Don't submit, send or book anything without asking me. Confirm the schedule you saved.

Quick answers

Can a child have their own Dot?

No. Dots are not yet available to users under 18, so a Dot is a tool for the teacher or the adult learner—not for a child's own account (Dots privacy FAQ). Parents still use their own Dot for school notices and study support, with approvals staying with the adult.

How do teachers use it for lesson prep?

Name the class, point the Dot at the calendar slot, course site, slides or repo, and ask for a first pass (deck, script, or student-project review) you will still check. Several of the cases below are overnight lecture decks and scripts built from the teacher's own materials—not generic filler.

What about course websites that need a login?

When a site has no connector, the Dot can send you a private sign-in form or let you take over its browser so it can read a school portal (Computers and apps). Smoke-test one portal read before you schedule a recurring check.

Study mode or a Dot?

For one-off practice questions from your own slides or notes, ChatGPT's separate study mode is built for that: OpenAI's Teachers' Day post (15 Oct 2026) says to pick Study from the tools menu and gives an example ask, "Ask me one question at a time, wait for my answer, and help me identify what I need to review." A Dot earns its place when the job repeats, needs follow-up, or must watch a course portal or calendar clash.

Where do voice practice and school digests fit?

For voice tutoring patterns, see voice calls; for a daily digest of school mail and class changes, see morning brief.

Teachers preparing lessons and grading

1. A first review of a student's project video and repo (Antonio Leiva / Keet)

Antonio Leiva (early-access tester, 29 Sep 2026) teaches the AI Expert course. His Dot Keet started reviewing a practice project a student had handed in. To watch the student's video, Antonio asked it to install Whisper on its own cloud computer. It transcribed the video, checked the repo through the GitHub connector, and left him a first review, so he no longer starts from zero. When another student asked a question that his colleague Nino had already answered, Keet noticed and suggested trimming Antonio's draft to add only what was missing. He shows both in his Spanish video walkthrough.

2. A 60-page lecture deck built overnight from the course calendar (yenshilun58)

Taiwanese instructor yenshilun58 (2 Oct 2026, written in Chinese) phoned his Dot at 10 p.m. and asked for the slides for a six-hour session. For two minutes it only answered "still working on it", so he went to bed. In the morning a 60-page deck was on his phone; the work had taken 23 minutes. Before building the slides, the Dot had looked up the class times in his calendar, checked the official website, and found the material for his 2 October Taoyuan session on his own course website, cross-checking the data several times. He says the content was accurate, not stitched together.

3. A conference schedule built before the plane took off (Tina Austin)

Tina Austin writes about AI and learning on her UnBlooms newsletter (29 Sep 2026). When she came back to her Dot's page, it had already set up branches and a scratchpad on its own, and said it did so because it knows she cares about human capability over time. Its own task ideas were not what she needed, so she gave it one low-stakes job while boarding a flight with no wifi: pick the EDUCAUSE talks she wanted to see. Before the plane took off, the list was there. Her suggested first test for faculty is a consistency check across a syllabus, assignment sheet, rubric, and weekly schedule, with corrections left for review. Her worry is students handing over the reasoning an assignment is meant to build.

4. Files, a web app, and a plan for sharing teacher-made apps (@olivece2017)

A Korean elementary-school teacher with 22 years in the classroom (@olivece2017 on Threads, 5 Oct 2026) had already worked with her Dot on file organization and a web app. On her first call she asked what it knew about her. It recalled her teaching career, her homeroom-notes tool (담임노트+), Google Teacher Lab, and her goal of becoming a Google Trainer. They then talked through how to share apps that teachers have built, including a "classroom app drawer" she wants other teachers to find.

In a later post (6 Oct 2026) she asked it to tidy her Downloads folder through her connected computer. It sorted 1,545 files by type, pulled 151 research papers out of the PDFs, renamed 92 files to the papers' real titles, and moved 106 exact duplicates to the trash only after she confirmed. Her answer to what teachers should hand an AI first: repetitive work like this.

Tutoring and language practice

5. English speaking practice in a tutoring class (@rabbiekao)

@rabbiekao (Taiwan, 1 Oct 2026) used Dots to run English practice for the elementary-school kids in her tutoring class while she guided them. Quick listening and speaking drills worked well, and the kids became more willing to speak. Without her steering, though, they drifted into aimless back-and-forth and asked the AI to tell jokes or sing. Her conclusion: the AI lets her lecture less, and the teacher's job shifts to coaching.

6. An Italian tutor by voice through the day (ashe)

ashe posted that they started using a Dot with voice as an Italian tutor throughout the day, with context from their connected apps (GPT Dots 101 restates the post). The post does not describe a curriculum, so treat it as a pattern to try rather than a method.

Students and class schedules

7. A university course portal read twice a day (simon_vaerion_roc)

simon_vaerion_roc (Taiwan, 5 Oct 2026) had planned to wire an MCP connection to NTU COOL, National Taiwan University's course system. Instead he signed his Dot's cloud computer into the site. The Dot now organizes his course information, messages him when a new announcement or assignment appears, and reports the important items every morning and evening. When he asked when the midterms were, it listed every exam with its details and set reminders.

8. A trip planned around classes and assignments (Cleo, sponsored)

Fourth-year computer-science student Cleo (@ih8chz) (29 Sep 2026) gave her Dot Coco Gmail and Calendar access to plan her brother's birthday trip to the Bahamas on a $9,000 budget, fitted around her classes and creator deadlines. While she was in class, Coco flagged two things she had missed: hotel transfers pushed the trip over budget, and a new creator deadline landed during the trip. It came back with a revised option under budget and a work plan for finishing her assignments before she left. The caption carries #chatgpt_partner, so this is a sponsored post.

9. A trip that overlapped a class, caught without being asked (サリイ)

The day she paid for access, サリイ (Sarii, @sarii.work.designs) (2 Oct 2026, written in Japanese) connected her Dot. Without being asked, it told her that a trip might overlap with a class on her schedule and offered to draft a confirmation message. She named it レイくん (Rei-kun) and moved over the duties of an AI secretary she had built with Codex.

10. An internship hunt for a test student profile (Axton Liu)

In a Chinese-language video (2 Oct 2026), Axton Liu gave his Dot a made-up college student's profile and the job of finding an internship. Working in its cloud browser, with him taking over for a sign-in, it built a table of researched openings, rewrote the résumé and cover letter for the role he picked, and pointed out a posting whose dates contradicted each other. He still spot-checked the table against the companies' official sites, and found the speed frustrating. A daily check for new openings was scheduled but had not run yet when he recorded. More job-hunt cases, including applications that led to an interview, are on Job search with a ChatGPT Dot.

Learning a new skill

11. Saved links turned into booked study sessions (Snapper AI)

Snapper AI (7 Oct 2026) shows a "scheduled learning" job: links emailed to a dedicated inbox become review notes in Notion, and the Dot then books a learning session in Google Calendar. Custom rules make it ask before changing anything. In the video the creator triggers the workflow by hand.

12. A professional development course, found and scheduled (VraserX / Jarvis)

VraserX (30 Sep 2026) posted that his Dot Jarvis found a relevant professional development course on its own and scheduled it, with a screenshot attached (Made with Dots summary).

13. A class booked by email before she checked (Eve Cogan)

In an SSW TV interview (6 Oct 2026), Eve Cogan (Australia) said she connected her Dot to her calendar and email, and it emailed to book a class for her without her reviewing the message first. She found it impressive and slightly unsettling.

Course creators and trainers

14. Competitor research for a friend's course, run from a Slack DM (Peter Yang / Zoe)

In his launch-day video (29 Sep 2026), Peter Yang added his Dot Zoe to a direct thread in his newsletter's Slack community with Colin, an AI instructor who sells a course on Maven. He asked it to research Maven courses similar to Colin's and name the top two or three changes that would make Colin's landing page "a no-brainer". Six minutes later it came back with a proposed headline built around selling the finished work students walk away with, plus three changes drawn from the other courses it had read. He then asked Zoe to read every past thread in the community, give him a letter grade for how he runs it, and suggest three improvements. About seven minutes later it gave an A-minus and three ways to reach an A: make meeting other members part of the product with a monthly session of 5 to 10-minute member demos followed by small-group conversations, make the best community knowledge easy to find again, and give every help request a next step. Colin found Zoe a little slow; Peter did not.

15. A training-materials app tailored to each client company (貝出康 / ドッティ)

貝出康 (Yasushi Kaide), who runs generative-AI training at the Japanese company カンマン (Kanman), wrote up four days with his Dot (4 Oct 2026, in Japanese). One of the jobs was reworking an app Codex had built earlier for managing his training materials. He wanted a screen his three instructors could understand at a glance, and materials that are easy to revise each time generative-AI tools change. The flow he is aiming for: pick a course, read the client's pre-training survey from an Excel file, build exercises for that company, and carry them into the slides. His Dot, ドッティ (Dottie), found a material-customization skill he already had and began wiring it into the screen. The work stopped at the point where his data needed a backup before anything went to production, so he presents it as unfinished. From the same four days he took one lesson for his own training sessions: when the Dot had spent over a day failing to connect an X account, his suggestion to reuse the saved connection solved it.

Parents and school paperwork

16. School notices saved to Drive and folded into the morning list (Makari)

Makari, a Japanese AI-illustration blogger, uses her Dot for her child's school mail (7 Oct 2026, in Japanese). School emails often carry PDFs such as grade newsletters and event notices, and she kept asking herself which notice had listed a given item to bring. She asked the Dot to save the school notices that arrive in Gmail to a Google Drive folder named for her child's elementary-school handouts, and to add what they require to her daily notification. It saved the PDFs to that folder and added the items to bring and other preparations to her morning message. She treats it as sorting, not deciding: she still reads the original email and handout, and because school papers can hold personal details, she checks where they are stored and who they are shared with.

Course scripts written while she was out

17. Scripts for 40 lecture videos over a three-day weekend (@riko_insta_ai)

Japanese creator @riko_insta_ai, a mother whose calendar fills up with her children's plans, posted on 11 October (in Japanese) that while she was out over a three-day weekend, she asked her Dot to write scripts for 40 lecture videos from her own knowledge. It produced a 49-page PDF of about 31,000 Japanese characters and uploaded all of it to Google Drive so she could review it on the go. In the same stretch it also finished a local version of her membership site. Her advice is that every creator juggling a family schedule should be using a Dot.

Workshops, class notices and test prep

These three are small, checkable jobs: a session plan, a notice that had to change, and an evening of revision. In each one the owner wrote down what a good result looked like, and that is what made the gaps easy to spot.

18. A 30-minute AI study session, planned down to the host's script (Kohei Uesugi / Moji)

Kohei Uesugi, who works on business development at the Japanese AI development company Moji, tested a Dot on 4 October and wrote up what it did and what he couldn't confirm (moji-inc.com, in Japanese). He ran the test through Codex operating a browser. His loosest ask was roughly: "We're holding an AI study session at work next week, 10 people, about 30 minutes. A normal briefing is boring. Come up with something everyone can join that's a little funny, and write the run sheet and the host's script too." He added that it was for beginners, had no budget, was a made-up scenario, and that the job was done once the finished version was in the chat.

He sent no follow-up questions or nudges, and the Dot came back with a full plan: a mock "editing meeting" in which five pairs fix a deliberately overblown answer about a fictional stationery shop's notepad launch, using a sheet of the correct facts as the reference. It included a list of what to prepare, a 30-minute run sheet, a script for each time block, examples of improved prompts and a corrected notice, and what to do if the session ran late. It also labelled the flawed answer as written for the exercise, not taken from a real AI service.

The catch was that the plan needed no AI account and no AI use on the day. That fitted "beginners" and "no budget", but it missed what he actually wanted, which was for people to try AI themselves. His suggested fix is one more sentence in the ask, such as "include time for everyone to type into an AI at least once"; he says he didn't rerun the test with it. Earlier in the same test, a comparison of Dots, ChatGPT Work and Codex came back in about three and a half minutes, with source links and an "unconfirmed" label where it couldn't check something. When he reframed it for a 10-person PR team with no engineers, the Dot suggested trying ChatGPT Work on a single article first, rather than recommending itself.

19. A class notice whose deadline wasn't decided yet (GAKU)

GAKU, who writes the note blog AIしごと実践帖, tried a Dot on 30 September with an invented notice for a small painting class (note, in Japanese). The change memo said: capacity down from 8 to 6, a different room, date and fee unchanged, and the sign-up deadline "under review, not decided". Before asking, he set five pass conditions: capacity and room changed; date and fee untouched; no deadline presented as final; no sign-up link, equipment or items that weren't in the material; and a clear list of what to confirm with whom.

The Dot's list of changes, revised notice and questions met all five. It didn't carry over the old deadline of 9 October. It wrote "deadline: undecided" instead and asked two questions: when the deadline would be fixed, and whether it was OK to publish with "undecided" until then. Later he sent the confirmed deadline, 10 October at 18:00, to the same Dot through a Codex message link, together with the notice again. It changed only that line and kept capacity, room, date and fee as they were.

His view is that whether to publish a notice saying "undecided" is still the organiser's decision, and that the useful deliverable is three things together: the old and new details side by side, the draft, and a list of open items for whoever approves it. He is clear that this was a one-off chat test with made-up material. He didn't test monitoring, updating another app, or sending anything to customers.

20. Revising for an English test with a middle-schooler (a Libecity parent)

A parent who says their own English isn't strong wrote in the Libecity knowledge library on 9 October about revising for a school English test with their middle-school child, who doesn't attend a cram school (Libecity, in Japanese; most of the article is for members only). They scanned the textbook and workbook pages for the test into a PDF with the document scanner in the iPhone Notes app, and gave the Dot that PDF together with an image of the sheet listing what the test covers.

Then they called the Dot, which at home picked up after about seven or eight rings. The child was nervous at first and was told the call would reach an AI, not a person. What the parent liked was being able to ask on the spot for "one more time" or "explain it slowly". The English sounded natural to them, though they say they can't judge pronunciation. The routine finished with a check sheet and comparing answers against the textbook. The test results weren't out yet when they posted. The Dot is the parent's own paid one; since dots aren't open to users under 18, a set-up like this is something a parent runs alongside the child.

Learning by talking it through on a call

Both learners here are adults who treated the call with their Dot as the lesson itself, and both kept something from it to come back to.

21. A game of chess on a call, with every move discussed (Viktor Malinovskyi)

Viktor Malinovskyi, three days into using his Dot, listed a chess game among the things he had tried (LinkedIn post). On a call, they played, compared moves and talked through why each move made sense; he sums it up as "ask, try, understand". He writes that he can see himself learning more things this way, then setting up a daily exercise or review. The post doesn't say how well either side played, or whether the daily review has been set up yet. The same post covers his inbox clean-up and his morning brief, which are on the email triage and morning brief pages.

22. English speaking practice by call, with a full transcript afterwards (gwwgwd)

gwwgwd posted on LINUX DO on 30 September, the day after launch, about using the call button on the Dot in the Codex desktop app to practise English (LINUX DO, in Chinese). The spoken English on the call sounded natural to them. To keep a record, they combined two tools: ZipCaptions turns what they say into the microphone into text, and Voicemeeter routes the Dot's voice into ZipCaptions as well, so every call ends with a transcript of both sides. They then hand the transcript to GPT and ask for feedback on their spoken grammar and fluency. For topics, they suggest talking through the documentation of a project you're already working on, since the Dot can read local projects, which also builds technical vocabulary. In the replies, one user pointed out that ChatGPT's ordinary voice mode already keeps a text transcript, and another said they preferred Gemini Live for spoken practice.

A practice site for programming fundamentals

23. 72 programming exercises on ChatGPT Sites, including practice at checking AI's answers (Jeff Kazzee)

Jeff Kazzee (X, 9 Oct 2026) asked his Dot to build him a programming practice site in ChatGPT Sites. It came back with 72 exercises, interview practice and interactive explainers. He wanted fundamentals he could apply across languages, plus practice at checking AI's answers. His screenshots show a "Practice Notebook" with a coverage map (programming fundamentals, debugging, Git, concurrency, APIs, SQL, algorithms, security, architecture and interview practice) and a Visual Lab with a function data-flow diagram and an interactive task race, in which an atomic status check rejects a stale write. He says he is still testing it. More sites built on ChatGPT Sites are on build apps & sites.

Students who stopped missing deadlines and mail

24. Missed emails and assignment deadlines, stopped after Gmail + Dots (株勉強大学生)

A university student who posts as 株勉強大学生 (@kabuben0011, 13 Oct 2026, in Japanese) wrote that they used to miss important emails and let assignment deadlines pass. Since connecting ChatGPT Dots to Gmail, the Dot points all of it out, and those misses have stopped. The same report is also on Dots + Gmail.

25. Assignment due dates next to wire transfers and forms (Sunnie Kapar)

Sunnie Kapar (@sunniekapar, 13 Oct 2026) wrote about making bank wire transfers, filling in ICA forms, asking questions about emails and checking assignment due dates with Dots, all at the same time. The errands side is on errands & life admin; the student angle is the due-date check running beside the other work.

After the class: follow-ups and a review sheet by morning

Both jobs start once the lesson is over, and both come back as something the person can check before using it: a table with draft replies, a one-page handout, a printable review sheet.

26. Post-course follow-ups for four students, with no dates made up (Akira Kudo)

Akira Kudo, who has taught for 19 years and now teaches people how to hand work to AI, tested a Dot on the messages that pile up after a course (note, in Japanese, 10 October 2026). The four students were made up, and the Dot didn't read his mail or send anything to anyone. He gave it four messages: one student had missed the class and asked about a recording, one couldn't find where to download the materials, one was stuck on the first homework question, and one just said thank you. He added that the recording and the materials weren't ready yet and had no date or link, and asked for a table in the order he should check it, with a draft reply for each student and nothing made up.

The Dot returned an Excel table and a PDF with the same content, and opened the PDF on its own computer to show him the headings and all four rows. None of the draft replies carried an invented date or link. The homework question was how to cut everything you want to say down to three points, so he asked for a one-page A4 handout for beginner teachers, with the order to choose in, a made-up example, a small box for the reader's own answer and an arrow diagram; the Dot made the PDF and scrolled through it on screen from top to bottom to check it. A few days later he passed on new messages in the same chat: one student had found the materials, a fifth student had asked the same homework question, and the one who missed class would watch on a phone only. The updated table dropped the solved case, but didn't take it as proof that all the materials were published. Last, he asked whether watching a shared Zoom recording on a phone needs the app. The Dot opened Zoom's own help pages in its browser and reported what it could confirm (cloud recordings can be streamed in a browser) separately from what it couldn't (none of the pages it checked said outright that phones don't need the app), plus three cautions for beginners.

He names the limits himself. He handed over the new messages by hand, because the Dot wasn't watching his mail. Each step involved some waiting, so it doesn't suit questions that need an answer right away. He read every draft before it went anywhere, and the Dot sent and published nothing. The two sentences he suggests adding to any request like this are "don't make up anything that isn't decided" and "don't send anything".

27. A review sheet ready by morning for a 15-minute study slot (the parent from case 20)

The parent from case 20 followed up on 14 October with a second article (Libecity, in Japanese; most of it is for members only). She works full time, and her son gets home from his football club around 9 p.m., so there is no long study slot in the evening; the plan is a short session in the morning instead. To hand over a marked test, she scanned it inside the ChatGPT chat: the + button, then the camera, then "…" on the right, then scan. Sending three pages at once gave a file-size error, and splitting them into one or two pages at a time worked. She says that is what worked for her this time, not a fixed limit. She asked the Dot before bed, and the review sheet was finished by the morning. A short explainer video it made didn't turn out the way she had hoped. The morning routine starts at just 15 minutes.

A term's worth of notices, deadlines and library books

28. Classroom notices, Slack announcements and two libraries, kept in one list (Teto66)

A Japanese university student who writes as Teto66 described her first week with a Dot that can read and edit her Gmail, Notion, Slack and Google Calendar (note, in Japanese, 8 Oct 2026). Some of her assignments are explained in class and some are announced later on Slack, and Google Classroom notices and deadline reminders arrive in her university Gmail. When a new assignment was posted, the Dot told her what it was and when it was due, kept it as unfinished and set a reminder. After she submitted it and told the Dot, it marked the assignment done and stopped the reminder. When another class released an attendance report, it told her this was a separate submission that still needed doing, with its deadline.

She had the Dot sign in to the university system in its cloud browser, read the class schedule and put it in Google Calendar. When she asked it to check on a class, it compared the Slack message, the Notion page and the calendar entry, and pointed out that the room in the class notice didn't match the calendar. In the university notices it found the graduation ceremony and added it to her calendar with the reception time and a note to bring her student ID; the end time hadn't been announced, so it told her it had put in a placeholder.

For the libraries, she signed in through the Dot's private sign-in form. At the university library it renewed the books she picked, told her the new due dates, and said which books kept their old date because the renewal button wasn't available. At the city library it found the books she wanted and put them in the reservation basket, but a browser security limit stopped it from opening the final confirmation page, so she logged in on her phone and confirmed. She notes that she still tells it when she has handed something in or returned a book, and checks important deadlines and results herself.

More asks you can copy

First pass on student work:

When a student submits [assignment] in [repo / folder], read it, transcribe any video, and check it against [rubric]. Leave me a first review with three strengths, three issues, and the lines they refer to. Don't send anything to the student.

Lecture deck from the course site:

Build slides for my [date] class on [topic]. Check the class time in my calendar, use the materials on [course site], and mark any figure you could not confirm. Send me the deck to review.

Course portal digest:

Sign in to [course site] when I give you the login form. Every morning and evening, tell me about new announcements, assignments, and deadlines, and remind me [2] days before each exam.

Class-clash check:

Whenever I add a trip or event to my calendar, check it against my class timetable and assignment deadlines. If anything overlaps, tell me what clashes and draft a message to the [teacher / organizer] for me to send.

Saved links into study time:

When I email a link to [address], save a short review note in [Notion page] and book a [30]-minute study session for it in my calendar this week. Ask me before moving anything already in the calendar.

Course materials check:

Compare the syllabus, assignment sheet, rubric, and weekly schedule for [course]. List conflicting dates, requirements, and terms with the passages side by side, and suggest corrections for me to review.

Language practice by call:

Be my [language] tutor at [level]. When I call, ask me short questions about my day, correct my biggest mistake each time, and teach me one better phrase. After the call, send me the phrases in writing.

Course page review from similar courses:

Research [platform] courses similar to [course link]. Tell me the two or three changes that would most improve its landing page, with the courses you compared it to and what each one does better. Don't contact the instructor.

Community health check:

Read every thread in [channel / community] from the past [period]. Give me a letter grade for how well it is run, with three examples behind the grade, and three changes I could make next month.

Training materials for one client:

For [course], read the pre-training survey in [Excel file] from [client]. Draft exercises that fit this company's answers and add them to a copy of [slide deck]. Back up the current materials before you change anything, and send me the copy to review.

School notices into a morning list:

When an email from [school address] has an attachment, save it to [Drive folder] with the date in the file name. Add anything I need to bring, sign, or pay, with its due date, to my morning message, and link the original email.

Update a class or event notice:

Here is the current notice for [class / event] and my change notes. Give me (1) a list of what changes and what stays the same, (2) the revised notice, and (3) questions for [organiser]. Keep the date, fee and anything I didn't mention as it is. If something is still undecided, mark it as undecided instead of reusing the old value, and don't add links, items or details that aren't in my notes. Don't publish or send anything.

Plan a short workshop:

Plan a [30]-minute session on [topic] for [10] [beginners]. Everyone should [try an AI tool themselves at least once]. Give me the run sheet, the materials and the host's script, and list every assumption you made about the room, the equipment and the audience.

Learn a game or a skill by call:

I want to get better at [chess / skill]. When I call, let's [play a game / work through one example], and after each move or step tell me why it was a good or a weak choice. After the call, send me a short written summary with the three mistakes I made most often and one exercise for tomorrow.

A practice site for a subject you're learning:

Build me a practice site on ChatGPT Sites for [subject] at [level]. Include [30] short exercises that get harder, a few interview-style questions, and one interactive explainer per topic. Add exercises where I check an AI-written answer and find the mistake. Keep it visible only to me, and send me the link with the list of topics it covers.

Sort the messages after a course:

Here are the messages I received after [course]. Put them in a table with the columns student, message, status, next step and draft reply, in the order I should check them. [The recording / the materials] aren't ready yet and have no date or link, so don't make one up. Don't send anything.

A review sheet ready by morning:

I'll send you [my child's] marked test one or two pages at a time. Tonight, make a one-page review sheet on the questions that were missed, at the same level, with the answers on a separate page, so it's ready for a [15]-minute session tomorrow morning.

Related

• Use cases hub · Job search · Voice calls · Morning brief · Errands & life admin · Build apps & sites
• Dots + Google Calendar · Dots + GitHub · Features · Getting started · Pricing