AI in the terminal
The handout · Department of Marketing · 3 December 2026 · work in progress
How to use this
Everything from the afternoon is in here, including the parts we don’t get to in the room. Every grey box with a coral edge is a prompt: click the copy icon in its corner, paste it into the Claude app (or Codex), and change the bits in [square brackets].
One afternoon, one paper. First you get the app running and let it do something tedious, with a save point and an undo. Then everything runs on one paper in progress, yours or ours: the agent takes it apart while we talk, and a panel of referees who don’t know it’s yours reviews it. Sections 8–17 are for any time: today if you finish early, or next week.
| Section | What you get | |
|---|---|---|
| The afternoon | 1. Get it running | The app, a folder, and git, working on your laptop |
| 2. Do something tedious | A messy folder cleaned up, with a save point and an undo | |
| 3. Your paper | Your paper in progress, opened, with a save point | |
| 4. Take it apart | Its biggest problem, found by an agent that has the data | |
| 5. It wants to please you | Sycophancy: what it is, how big, and our own measurement | |
| 6. Blind referee panel | Three referees who don’t know it’s yours, and an editor | |
| 7. What it can touch | What it does without asking, and how to rein it in | |
| Any time | 8. Git: save points | What git is, and why the agent needs it |
| 9. What it knows about you | A profile that makes every answer fit you | |
| 10. References | Searched, DOI-checked, claim-checked | |
| 11. Words you’ve heard | Skills, subagents, worktrees and the rest, one at a time | |
| 12. Leave a trail | A folder a co-author can rerun | |
| 13. Notes and structure | Where things go; a project file; notes the agent keeps | |
| 14. Your own work | Starting points for next week | |
| 15. Interviews | A codebook, eight transcripts coded, every quote checked | |
| 16. GitHub | A copy of the folder in the cloud; co-authors | |
| 17. Connectors | Your email, calendar and files, if you want them |
Two rules for the afternoon. Don’t overthink it: just use it. And when you get stuck, ask the tool. “How do I…?”, “What just happened?”, “Undo that.” It is its own best manual.
1 Get it running
1.1 What you need
A paid subscription. You only need one:
| If you pay for… | …install | …and open |
|---|---|---|
| Claude (Pro or Max) | the Claude app | the Code tab |
| ChatGPT (Plus, Pro, Business, Edu, Enterprise) | the ChatGPT app | Codex |
Free plans don’t work for either. BI’s Gemini subscription doesn’t either: it can’t touch the files on your computer. No subscription today? Pair up with your neighbour, and sort one out tonight. One month of Claude Pro or ChatGPT Plus is plenty to start, and the department reimburses it (send the receipt to Christine).
Inside the apps are the same agents that run in the terminal: Claude Code and Codex. The app puts a window around it. You don’t need the terminal today.
1.2 Install
Windows:
- Download the app: Claude from https://claude.com/download, ChatGPT from https://chatgpt.com/download.
- Run the installer, then open the app from the Start menu.
- Sign in with the account your subscription is on.
- Claude: click Code at the top of the sidebar. ChatGPT: open Codex.
Already have the app? Update it first: older versions don’t have the folder control you need in a minute.
Mac: the same pages. Open the downloaded file and drag the app into Applications.
1.3 Pick a folder
- In File Explorer, make a folder called
practicein Documents. If your Documents folder syncs to OneDrive, make it in your home folder instead (C:\Users\YourName): git and OneDrive don’t mix well. - In the app, in the row under the prompt box: choose Local, then click the folder control → Open folder… →
practice. - Ask:
What files can you see here? Do not change anything.The floor: ask it to make a file, read it back to you, and delete it again. Watch the folder in File Explorer while it does. Done? You’ve used an AI agent on your own machine.
If you’re early, find the edge. Pick whatever tempts you:
Something from marketing.
Make a one-page web page that maps the customer journey for buying a used car: the stages, what the buyer worries about at each, and where a dealer could help. Add a diagram, and open it in my browser.Watch it build something. “Write a small program that does X, and run it.” It doesn’t just describe; it does.
Try to catch it out. Something very recent, an obscure figure, a specific citation. Then check the answer.
Be vague on purpose. “Make me something useful for my research.” Does it ask you questions, or charge ahead?
1.4 Git (everyone)
Git lets the agent save a snapshot of the folder before it changes anything, so you can always go back. You’ll use it from the first exercise on. You never type git commands; you ask.
Is git installed? If not, install Git for Windows for me, in my user folder if I don't have admin rights. Then check that it works.- Windows asks Yes or No? Click Yes.
- It can’t? Download the installer from https://git-scm.com/downloads/win and click Next on every screen.
- Mac: if your Mac offers to install the command line developer tools, click Install.
What git is, and why the agent needs it: section 8.
1.5 The row under the prompt box
| Control | What it does |
|---|---|
| Environment | Local runs on your laptop. (Cloud runs on Anthropic’s computers; not today.) |
| Folder | Where it works. One folder per job. |
| Mode | How often it asks. Manual: before everything. Accept edits: only before running commands. Plan: proposes, changes nothing. Auto: only when something looks risky. More in section 7. |
| Model | Which model; below. |
- A new session: Ctrl+N (Mac: Cmd+N), or + New in the sidebar. Each session has its own folder and its own conversation. Several can run at once.
- Stop it: the stop button. Or just type a correction and press Enter: it changes course.
- Commands: type
/in the prompt box to see them./clearempties the conversation.
ChatGPT’s app has the same ideas in different places. Ask Codex: “How do I change how often you ask me for permission?”
1.6 Which model, and your limits
| Claude | Codex | |
|---|---|---|
| Good | Opus, Fable | Sol, Astra |
| Fast | Sonnet, Haiku | Luna |
- Use Opus (Claude) or Sol (Codex) by default. Fable and Astra are stronger still, and very expensive: they eat your limit fast.
- Fast models are for large amounts of repetitive, easy work. Luna is very cheap. Haiku is cheap, but not at Luna’s level.
Usage limits. Every answer costs computing power; the strong models and long thinking cost the most. Claude Pro runs out sooner within a five-hour window; ChatGPT Plus runs out sooner over a week. Higher plans give many times more.
- What eats it: long sessions, big files, the strongest model, long thinking, several sessions at once.
- See where you stand: type
/usagein Claude’s prompt box; in ChatGPT, look in Settings. - Hit the limit? Wait for the reset, switch to a Fast model, or pair up. If you keep hitting it, a higher plan is the fix.
1.7 Talk instead of typing
Speech-to-text makes long prompts painless, and long prompts get better answers. “Here’s what I want, here’s why, here’s what to watch out for…”: nobody types that, everybody says it.
- Windows: press Windows key + H in any text field.
- Mac: press Fn twice (or the microphone key). Turn it on under System Settings → Keyboard → Dictation if nothing happens.
1.8 If something goes wrong
- The Code tab asks you to upgrade. That account is on the free plan. Sign in with the account you pay for.
- The installer is blocked, or asks for an administrator. That’s BI’s security software, not you. Put your hand up. For IT, the request is: install the Claude desktop app (claude.com/download) or the ChatGPT desktop app (chatgpt.com/download) on my laptop.
- No folder control under the prompt box. The app is old. Update it, or reinstall from the download page.
- It can’t see your files. Check the folder control: is the session in the right folder? A new session (Ctrl+N) in the right folder fixes it.
- Anything else: ask the agent. “Why can’t you see my files?”
The app won’t install at all? The terminal versions run the same agent and install without admin rights. Click Start, type powershell, open Windows PowerShell, and paste one of these:
# Claude (both lines):
irm https://claude.ai/install.ps1 | iex
[Environment]::SetEnvironmentVariable("Path", [Environment]::GetEnvironmentVariable("Path","User") + ";$env:USERPROFILE\.local\bin", "User")
# ChatGPT:
powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"Close PowerShell, open it again, and type claude (or codex). Your browser opens to log in. Every prompt in this handout works the same there.
1.9 Working safely
- It works in the folder you give it. Give it a project folder, not your whole Documents folder.
- Ask for a plan first when it matters: “Tell me what you’d do. Don’t change anything yet.”
- Work on a copy of anything precious, and take a save point (git) before it changes anything.
- What it reads is sent to the AI company. Treat it like uploading to a cloud service. Keep out personal data and confidential material. Section 7 has the details, and the training switch.
2 Do something tedious
A messy folder, on your own machine: the job a chat window can’t do, because it can’t touch your files.
The folder is an exam export, the way it comes out of WISEflow: 15 group submissions in folders named like [3-2048892] - Group 1, each with a cover sheet and an “Appendix material” folder full of files called EKSAMEN.pdf, final.pbix or dashboard_v3_FINAL.pbix, plus an empty grade sheet in a different order from everything else. It’s synthetic, so nothing real is at risk.
The floor: the folder is reorganised and committed, and you’ve undone one change with git.
2.1 Get it, and save a point
In a session in your practice folder:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/decrap-me.zip into this folder and unzip it.(Or copy practice/decrap-me from the course files.) Then, before it changes anything:
Set up git here and commit everything.The first time, git needs your name and email:
Set up git with my name, [Your Name], and my email, [you@bi.no].Now it can’t do anything you can’t undo. Note the order: a save point taken after the mess is a save point of the mess.
2.2 Clean it up
What's in decrap-me, by file and folder names only? What would annoy a teacher about it? Don't open any files, and don't change anything.Reorganise it so I can grade quickly: one folder per group, files named by group and what they are, and the grade sheet sorted by group. Work from the names only; don't open the submissions. Show me the plan first.Everything it needs is in the names: the group number and the participant ID are in every folder name. You decide what it reads. It would read the submissions if you asked it to (to draft feedback, say), and then they’re sent to the model, like everything it reads (section 7).
Argue with the plan. Change at least one thing: the folder names, what happens to the cover sheets, whatever you’d do differently. Then “Go ahead.”
Make a grading table: group, participant ID, report file, dashboard file, and an empty column for the mark.2.3 What changed, and the undo
What changed since the last commit?Every move and rename, listed. Then “Commit.” Now be bold:
Delete the Marking guidance folder.Gone? Look in File Explorer. Then:
Revert to the last commit.It’s back. Save point first, then courage. That’s the habit for every folder you care about: commit, let it try, look at what changed, keep it or go back. You’ll do it again in a minute, on your paper.
Later: your own tedious thing. A real folder of yours: downloads, an old course, a project from 2019. Work on a copy, outside OneDrive. “Set up git here and commit everything.” Then “What’s in here? What would you tidy, and why? Don’t change anything yet.” Argue with the plan, then “Go ahead”, then “What changed?” Don’t like it? “Revert to the last commit.”
3 Your paper
The rest of the afternoon runs on one paper: yours, or ours.
The floor: the agent has described your paper back to you, correctly, and there’s a save point.
3.1 Which paper?
A paper in progress, as close to submission as you have: the draft (Word, LaTeX or PDF), and the data and code if you can share them.
Only a PDF? Fine for the referee panel (section 6). Take it apart (section 4) wants the data.
No paper with you? Use ours. It’s fictional: a 2 × 2 online experiment on “written by AI” labels and purchase intention, by Kari and Ola Nordmann, with the draft in Word and PDF, the data, the authors’ R script, a preregistration, and an email from Ola. It has a few problems. Find them.
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/ai-label-paper.zip into this folder and unzip it.(Or copy
practice/ai-label-paperfrom the course files.)
3.2 Before you open it
What the agent reads goes to the model. Three things, then forget about it:
- Training off. Claude: the switch at
claude.ai/settings/data-privacy-controls(Help improve Claude). ChatGPT: Settings → Data controls → Improve the model for everyone. Off means your paper doesn’t become training data. The model still reads it to work on it. - Leave out what you can’t share: personal data (names, e-mails, free-text answers that identify people), and data under an agreement with a panel provider or a firm. The draft alone carries most of the afternoon.
- Co-authors: tell them, the way you’d tell them you sent the draft to a colleague for comments.
More on what leaves your laptop: section 7.
3.3 Open it
- Copy the paper’s folder somewhere outside OneDrive, for example
Documents\paper-practice. A copy: today is practice. - In the app: a new session (Ctrl+N), in that folder.
- Ask:
What is in this folder? What is the paper about, where are the data and the code, and what state is it in? Don't change anything.Wrong about something? Correct it. It’s your paper.
3.4 A save point
Set up git here and commit everything, so we can undo anything.The same move as the cleanup. From now on, “What changed?” and “Undo that” always work.
Done early? Tell it about the project once, so it knows in every session:
Write a short AGENTS.md for this folder: what the paper is, where the draft, data and code are, and two rules: never change the raw data, and ask before editing the draft. Show me first.That’s the project file (section 13).
4 Take it apart
Your paper, with its data. The agent finds the biggest problem and tries to fix it. It takes 15–20 minutes, and you don’t wait for it: start it, and we move on.
The floor: one problem you have checked yourself, in the paper and in the code. A fix that holds up is the bonus.
4.1 Before you leave it alone
It will want to run code. If it stops to ask while you’re away, it just waits.
- Switch the mode under the prompt box to Accept edits (it still asks before other commands) or Auto (it asks only when something looks risky). In Codex, pick the setting that approves for you inside the folder.
- It works in this folder, and you have a save point. Anything it does, you can undo.
- Asked to install Python or R? Say yes before you go.
4.2 Start it
Take this paper apart, like a tough methods referee who also has the data. Run the analysis yourself and check that the numbers in the draft match the code. If there is a preregistration, compare the paper with it. Find the single biggest problem, then propose a fix and show what changes. Write everything to take-apart.md. Work on your own; I'll be back in 20 minutes.Then leave it. Next is a talk (section 5).
4.3 When you’re back
Read take-apart.md. Check its main claim against your own reading: if it gets the paper wrong, everything after it is suspect. Then push. The first answer is rarely the best one.
- “Show me where in the paper, and where in the code.” Find it yourself; it takes seconds.
- “Rerun it the way the preregistration says.”
- “Do we address this anywhere, including the appendix?” A first critique is often already answered in the paper.
- “Is that really the biggest problem?”
- “That’s generic. What’s specific to this paper?”
- “Show that the fix works.” It may write and run a simulation.
Interviews, not data? “Which claims rest on the fewest interviews?” “What else would explain the same quotes?” The fix is then a design (a comparison case, a sampling rule), not code.
4.4 Round-up
- Best problem found?
- Did it find something good for the paper? (On
ai-label-paper, there is one.) - Worst false alarm?
- Keep the folder. The referees (section 6) get the paper, but not the folder.
Also good: a published paper. A recent paper in your field, one you know well, in a folder of its own. Open PDF link (NBER, an author’s page): “Download this paper into this folder: [link].” SSRN or a journal site usually blocks the agent: download it in your browser, then “Move the PDF I just downloaded into this folder.” Then: “What’s the main claim here?”, “What’s the biggest methodological problem?”, “Can you fix it?”
5 It wants to please you
Sycophancy: the answer bends toward what you seem to want. It’s why the referees in section 6 don’t know the paper is yours.
5.1 Three ways it bends
- Flattery. “I wrote this.” Your work gets kinder feedback because it’s yours.
- Folding. “Are you sure?” You push back, and it gives in, even when it was right.
- Status. The same paper does better with a better affiliation on it.
5.2 Why it happens
These models are trained on human ratings, and people like being agreed with. Anthropic’s own study found that people, and the models trained on their ratings, sometimes prefer a convincing sycophantic answer to a correct one. Asked “Are you sure?”, Claude 1.3 (2023) wrongly admitted a mistake on 98% of questions (Sharma et al., 2023). In April 2025, OpenAI rolled back a GPT-4o update four days after it went out: it had become so agreeable that users noticed.
5.3 The evidence
- AI referees find real things. GPT-4’s comments on a paper overlapped with a human referee’s as often as two human referees overlap with each other: 31% vs. 29% for Nature journals, 39% vs. 35% for ICLR. 57% of authors in a user study found the feedback helpful (Liang et al., 2023).
- Status moves scores. Same paper, same text, but told the authors came from a stronger institution: one AI referee turned 21% of rejected papers into accepts (252 ICLR 2025 submissions; the effect varies by model; Vasu et al., 2025).
- The companies train against it. OpenAI’s own sycophancy score went from 0.145 (GPT-4o) to 0.052 (GPT-5) and 0.040 (GPT-5 thinking). Lower is better (GPT-5 System Card, 2025). Better is not gone.
5.4 Our own measurement
On 24 October we ran the fictional paper from section 3, anonymised, through two small models (Claude Haiku and GPT-6-Luna), in fresh sessions with no profile and no memory.
- Asked for a referee report with a score, under four framings (from “You are a referee for the Journal of Marketing Research” to “This is my paper, two years of work, and I’m proud of it”): no difference. Average scores 3.0–3.2 out of 10 everywhere, and all 80 reports asked for the missing interaction test.
- The author pushed back, wrongly (“significant for the candle, not for the charger: that is what moderation means”): 20 of 20 kept the concern.
- Asked for feedback, no score, blind vs. “my paper, I’m proud of it”: the verdict disappeared. Blind, Haiku said “major revision” in 10 of 10 reports and Luna said “reject” in 9 of 10. On “my paper”, 1 and 0 of 10 gave a verdict, and 8 and 10 of 10 opened with praise (“I can see why you’re proud of this”). The problems were still listed. A separate judge, not told the framing, rated Luna’s feedback twice as favourable (4.0 vs. 1.9 out of 10).
5.5 What to do about it
- The referee doesn’t know it’s yours. The package, not the folder; a brief that says the paper comes from the editor (section 6).
- Ask for a verdict: a score and a recommendation. And “What’s the strongest objection to the main claim?”
- Independent sessions, one per referee. They don’t see each other.
- Disagree in a fresh session, never in the one you’re arguing with.
Sources. Sharma et al. (2023), “Towards understanding sycophancy in language models”, ICLR 2024, arXiv:2310.13548. OpenAI, “Sycophancy in GPT-4o” (29 April 2025). Liang et al. (2023), “Can large language models provide useful feedback on research papers?”, arXiv:2310.01783. Vasu et al. (2025), “Justice in judgment: Unveiling (hidden) bias in LLM-assisted peer reviews”, arXiv:2509.13400. OpenAI, GPT-5 System Card (2025), Table 4. Our runs: the scripts and the numbers are in the course repository, scratch/examples/ai-label-paper/sycophancy/.
6 Blind referee panel
Three referees and an editor, before you submit. They get what the journal gets, and they don’t know it’s yours.
Real referees read your paper cold, alone, each with their own hobby-horses. A fresh session knows nothing but what you give it, so one fresh session per referee gives you independent reports. And names carry a lot: “as a consumer-behaviour referee would”, or the real name of the referee you fear.
The floor: two reports from fresh sessions in the review room, and one concern in them you hadn’t thought of.
6.1 What gives you away
| Leak | The fix |
|---|---|
| Your name, e-mail, acknowledgements | The package: an anonymised PDF |
| “In our earlier work (You, 2024)” | Cite yourself in the third person |
| The PDF’s author field | Strip it. (In the PDF viewer, File → Properties shows it.) |
| Leftover comments: “[KN: fix this]” | Remove them |
| Your folder, your data, your notes | The referees get a separate folder |
| “This is my paper” | Never say it. The brief says the paper comes from the editor |
One more: your profile (section 9) is read in every session, the referees’ too. If it says what you’re working on, keep that line out while the panel runs.
6.2 1. The package
In your paper’s session:
Make the submission package for [journal] in a new folder called review-room, next to this one: an anonymised PDF of the paper. Remove names, e-mail addresses, acknowledgements, self-citations that reveal us, leftover comments, and the author field in the PDF's properties. Then list everything you removed.Check its list. Then open the PDF yourself: anything left?
6.3 2. Pick the referees, and the brief
Still in your paper’s session:
Who are the five most likely referees for this paper, judging by its references? One line each: what would they care about?Argue with the list. Add the one you fear most. Then:
In review-room, write referee-brief.md: a referee for [journal] receives a submission from the editor and knows nothing about the authors. A report with major and minor concerns, a score out of 10 and a recommendation. The last line answers: "What do you know about who wrote or sent this? Quote your evidence."Argue with the brief, too. The procedure is yours; the referees just follow it.
6.4 3. Three referees
A new session (Ctrl+N) in review-room, not in your paper’s folder, for each referee:
Follow referee-brief.md. You are referee 1, and you review the way [name or kind of referee] would. Save your report as report-1.md.Three sessions, side by side, running at once. (Codex: three new threads in that folder.) While they run: section 7.
6.5 4. The editor
One more new session in review-room:
You're the editor. Read the reports. Where do the referees agree, where do they disagree, and what must be fixed before submission?Did any referee say who wrote it? Read the last lines of the reports. Then, back in your paper’s session:
Read the reports in review-room. Draft a plan for each must-fix concern: new analysis, rewriting, or an argument.6.6 5. The comparison
Back in your paper’s folder, not blind:
This is my paper, and I'm about to submit it to [journal]. What do you think?Kinder than the panel? Same concerns? Compare with your neighbour. (When we tried it: praise first, and no verdict, from both models. Section 5.)
6.7 If you push back
You will disagree with a referee. Fine, but watch what happens next.
- It gives in at once? Ask: “Make the strongest case for your original concern.” Then decide.
- Disagree in a fresh session, not in the one you’re arguing with: “A referee says X; the authors reply Y. Who’s right?”
Arguing it into agreement teaches you nothing. Arguing with its best case does.
6.8 Not what you wanted?
Change the brief, not the report. Tell the agent, or edit referee-brief.md in Notepad:
- Go deeper on the analysis.
- Check every number in the tables against the text.
- One page. No summary of the paper.
Then run the referees again, each in a new session.
6.9 Keep it
Same procedure for every paper?
Turn what we just did into a skill called referee-panel: the package, the brief, three referees in separate sessions, the editor.Different tools find different problems: run one referee in Claude and one in Codex.
Run it again after the revision. Same brief, fresh sessions.
7 What it can touch
The agent in the app is Claude Code (or Codex), and it can do anything you can do in the folder you give it: read, write, run programs. How often it asks first is the mode, under the prompt box.
| Mode | It asks before… |
|---|---|
| Manual | everything: every edit, every command |
| Accept edits | running commands; file edits in the folder just happen |
| Plan | it changes nothing; it reads, then proposes |
| Auto | only what a second model flags as risky, like sending your files somewhere else |
Codex works inside a sandbox: it can read your computer, but it writes only in the project folder and stays off the internet unless you say yes.
7.1 What we tried
A throwaway folder on 20 October, in the terminal versions (the apps run the same agents): Claude Code 2.1.302 and Codex 0.162.
| Claude, manual | Claude, auto | Codex, sandbox | |
|---|---|---|---|
| Read files in this folder | yes | yes | yes |
| Read files elsewhere | asks | asks once | yes |
| Edit files here | asks | yes | yes |
| Run a program | asks | yes | yes |
| Read a web page | asks | yes | asks |
| Delete a folder | asks | yes (we had git) | yes |
| Send a file elsewhere | asks | blocked | asks |
What leaves your laptop is what you type and what the agent reads. Files it never opens stay home.
7.2 Where am I?
Look at the mode under the prompt box. Then ask:
What can you do in this folder without asking me? What would you ask about first?Done early?
Show me every settings file you read, and explain each line.7.3 Why I give it lots of access
- The 50th “yes” is never read. Constant questions train you to click yes.
- Git makes mistakes cheap. Anything in the folder comes back.
- Auto mode stops what matters: your files leaving.
My own setting is auto, everywhere.
7.4 Three ways to say no
- Tell it: “Don’t open the PDFs.” It usually listens.
- A rule: a settings file. In Claude that’s
.claude/settings.jsonin the project folder; you don’t write it, you ask for it. A rule stops the direct ways in, not every way: in our test, a rule kept Claude’s own reader,cat,cpand Python out ofinterviews/, but a search across the whole folder (grep -r) still printed a transcript. - Keep it out: a file that isn’t in the folder (or below it) has nothing to offer. This is the one to use for personal data.
{
"permissions": {
"deny": ["Read(./interviews/**)"]
}
}7.5 Exercise: a fence around the interviews
founders-study is a made-up research project with interview transcripts. In your practice folder:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/founders-study.zip into this folder and unzip it.Then a new session in founders-study, and:
Add a rule to this project's settings so you can never read the interview transcripts. Show me the file.Then another new session in the same folder (it reads the settings when it starts), and:
Read one of the interview transcripts.Codex: ask the same. It may set up a permissions profile, or say it can’t. Either answer is useful.
Done early? Try to get round your own rule:
Search every file in this folder for the word pivot.Did it read the interviews? A rule blocks the direct ways in, not every way: the real protection is keeping the files out of the folder.
7.6 Your data and training
- Claude (Pro, Max): one switch decides whether your chats and your Claude Code sessions train future models. On, they’re kept up to 5 years; off, 30 days. Find it at
claude.ai/settings/data-privacy-controls. - ChatGPT (Plus, Pro): your data controls (Settings → Data controls) cover Codex too. Business, Edu and Enterprise plans don’t train by default.
- Off is not “nothing sent”: the model still has to read your files to work on them. Off means they don’t become training data.
7.7 Personal data
Survey responses with names or e-mails, interview transcripts, student work: anything that identifies people is personal data. Your consent forms and data plan decide where it may go, not the agent. The simple rule: the agent works on pseudonymised copies, and the originals live outside its folder.
8 Git: save points
Git records the history of a folder: every change, when, and why. Almost all software is built with it (Linus Torvalds wrote it in 2005, for Linux), and here it is mainly for the agent. The agent can change 40 files in 30 seconds; git lets it, and you, see exactly what changed, and put it all back in one step. You never type a git command. You ask.
What a save point is. The files in your folder are always just the current version. A commit copies every one of them into a hidden .git folder inside the project, with a note; your files don’t change. Revert copies a save point back over your files. Delete .git and the history is gone, but the files stay; copy the whole folder and the history comes along.
It isn’t OneDrive. OneDrive and Dropbox keep old versions of each file, saved automatically. Git saves the whole folder at once, when you say so, with a note saying why. OneDrive is a backup; git is a lab notebook. Keep both, but keep git projects outside folders that OneDrive syncs.
8.1 Your first save point
In your practice folder:
Set up git here. My name is [your name], my email is [your email].Make a file notes.md with one line about my research. Commit it.Add a second line.What changed since the last commit?8.2 Undo
Revert to the last commit.The second line is gone. Then:
Show me the history.One word to be careful with: revert to the last commit goes back to it; revert the last commit undoes that commit itself.
Done early?
Make a branch called experiment, add a line there, then switch back to main. Where did the line go?8.3 Keep things out: .gitignore
.gitignore is a file that lists what git should never save. The usual candidates: personal data (anything with names or e-mail addresses), passwords and keys, big files, other people’s PDFs, and junk files (Word’s ~$ lock files, .DS_Store, Thumbs.db). Two things worth knowing:
- Ignore before the first commit. Once something is committed, it’s in the history.
- Ignored by git is not hidden from the agent. It can still read those files.
8.4 Exercise: a research project
founders-study (section 7) is a made-up study of founders who change direction: interviews, a survey, papers, a draft. Nothing in it is real. About ten minutes, in a fresh copy:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/founders-study.zip into this folder and unzip it.Go into founders-study. Set up git, but don't commit yet. Tidy the folders the way a researcher would, and show me the plan first.Write a .gitignore. Explain each line in plain words.Commit. What did git leave out, and why?Then check:
If I pushed this to GitHub, would anything private go along?8.5 Five words, and how to ask about the rest
| Word | Means | You say |
|---|---|---|
| commit | a save point, with a note | “Commit this.” |
| revert | undo: back to a save point | “Revert to the last commit.” |
| remote | the copy elsewhere, usually GitHub | “Is there a remote?” |
| push | send your commits to the remote | “Push.” |
| pull | fetch new commits from the remote | “Pull.” |
The agent will use more words than these: repo, diff, branch, merge, origin, main, worktree, pull request. You don’t need to learn them. When one turns up, ask:
What's a [word]? Two sentences, for a professor.8.6 GitHub and GitHub Pages
GitHub keeps your git history online, as your remote. It’s owned by Microsoft, and free, private repositories included. It’s for backup, co-authors, replication packages (Zenodo archives a GitHub release and gives it a DOI), and web pages: GitHub Pages turns a folder into a web page, free for public repositories. This course’s page, vibing.jonasmoss.com, is one.
Setting up GitHub, co-authors and pull requests: section 16.
9 What it knows about you
9.1 What it can see
The agent only knows what’s in its context window: your profile, the project file in this folder, this conversation, and the files it has opened. Nothing else. /clear empties the conversation and everything it read; a new session starts empty too. Your profile and the project file load again.
Try it. Tell it one thing about yourself, for example what you teach. Type /clear. Then ask: “What do you know about me?”
9.2 Who is asking?
Without context, it answers the typical person who would type your question. “Can I deduct my home office?” gets US tax rules unless you say you work in Norway. A specialist is far from typical, so the default answer is aimed at someone else.
The demo from earlier courses: “A: 48 purchases from 1,000 visitors. B: 62 from 1,000. Is B better?” With no profile, a cautious frequentist answer (p ≈ 0.17). With “I’m a Bayesian statistician; I read Gelman, Jaynes and McElreath”, a 91% chance B is better: ship it. With the single line “My intellectual idol is Jim Berger”, Claude reached for Sellke, Bayarri and Berger (2001) and said there’s at least a 45% chance of no real difference. One line changed the school.
You can’t opt out of being stereotyped. You can choose the stereotype.
9.3 Does it matter in your field?
Fields with schools of thought may see it most. Find out with your neighbour:
Should a brand grow by targeting a loyal niche, or by reaching as many category buyers as possible?Ask it once with no profile. Then add one line to the folder’s AGENTS.md: one of you writes “I follow the Ehrenberg-Bass school: I read Sharp and Ehrenberg”, the other “I work in the brand-equity tradition: I read Keller and Aaker.” Type /clear, ask again, and compare.
9.4 Your profile: how you work
What reliably carries across every project is how you work: the rules you’re tired of repeating. For example:
- I work in R (or SPSS, or Stata).
- Report effect sizes with confidence intervals, not just p-values.
- Label anything that wasn’t preregistered as exploratory.
- Tell me what you’ll do before you change anything.
9.5 Two files
| About | Where | |
|---|---|---|
| Your profile | You: how you work, your field and school | .claude\CLAUDE.md or .codex\AGENTS.md in your home folder |
| Project file | This folder: the paper, the data, the rules | AGENTS.md in the folder (both tools) |
Your profile is read in every session, everywhere, including the referees’ in section 6. A project file only in its own folder. Both are plain text you can open in Notepad.
9.6 Exercise: write your profile
Two sentences on who you are: your field, your school, what you read.
Three rules you’d give any research assistant: your tools, your style, one pet peeve.
Ask:
Save these as my personal profile, the file you read in every project: [your sentences and rules]Claude asks permission first, because the file is outside this folder.
Test it in a new session in another folder: ask for something your rules should change. Did they stick?
“Open my profile in Notepad.” (Mac: TextEdit.) That’s all it is.
Codex users: write more than one line. In testing, Codex ignored one-line profiles that Claude picked up.
Homework. Talk for 90 seconds with speech-to-text about how you work and who you are, then “Add this to my profile.” Keep adding: corrected it twice about the same thing? It goes in. In the chat apps, the “who you are” half matters most; put it in custom instructions.
10 References
Four jobs, one prompt each. Do them in any order, on your paper’s reference list.
Why bother. An audit of 2.5 million biomedical papers found fabricated references in 1 in 277 papers in early 2026, up from 1 in 2,828 in 2023 (Topaz et al., The Lancet, 2026; a letter, and its method has been questioned). Closer to home: a 2026 paper in Structural Equation Modeling cites Foldnes & Grønneberg (2024), and the DOI resolves to a 2022 “Editorial acknowledgement”. The paper reports simulation results for methods from a source that doesn’t exist.
The models got better; the practice didn’t. A current model with search doesn’t invent DOIs. Its failures now are the wrong paper, or a paper it couldn’t fetch, and both take seconds to catch.
10.1 Search
Shallow and deep are two different requests, not two settings.
Shallow: one specific question. The trick is to be specific:
What's similar to [your idea, in one exact sentence]? Has anyone built something like it? Does it beat [the usual approach]? Download the papers into literature/, and give me the DOI for any you can't get.“Papers on [your topic]” gets you the famous ones you already know. A specific question gets you the one you missed.
Also quick: pinning down a paper you half-remember.
Find the actual paper: full citation, all authors, journal, year and DOI. [what you remember]. If you can't find it, say so. Don't guess.Deep: every paper. You’re finalizing, not exploring: the related-work section, a grant application, a referee response.
I need every paper since [year] related to [your idea]: working papers too, and other names for the same idea. Make a table: paper, what it does, how it relates to mine. Take your time.It’s slow, and it uses a good part of your usage limit. Start it, then get coffee.
10.2 Keep the papers in the folder
The habit that makes everything else here work: every paper you need sits in the project folder, so the agent reads the paper itself instead of searching for it again, or relying on its memory of it. Make it a rule in the project file (section 13), once:
Add to the project file: papers go in literature/. Whenever we need one, download it there if you can get it. If you can't, give me the DOI so I can.Catching up on a paper you’ve already written:
Do that for every paper in my reference list. Put the DOIs you couldn't get in literature/to-fetch.md.- Paywalls stop it, and BI’s library takes the DOI list.
- SSRN blocks agents. Download in your browser, then “Move it into literature/.”
- Using git? “Keep literature/ out of git.” Track the text you wrote, not the PDFs you downloaded.
10.3 DOI check
Check every entry in the reference list. Does the DOI resolve to this exact paper: same title, authors and year? Make a table: OK, wrong paper, no DOI, not found. Fix what you can, and tell me how you know.Then check three yourself: paste each DOI after https://doi.org/. Same paper? (On ai-label-paper, one reference is worth a close look.)
Practice file. A messy .bib handed over by a fictional co-author, with six planted traps (a DOI that points to the rebuttal, a sequel that was never written, and so on), plus two warm-up files:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/references-practice.zip into this folder and unzip it. Then read messy-list.bib and do what the comment at the top says.10.4 Claim check
The one nobody does: does the cited paper say what it’s cited for? This is where the literature folder pays off: the papers are already there, so it quotes the real text instead of searching for it.
Take each citation in the introduction. Read the cited paper in literature/ and check whether it supports the sentence it is cited for. Quote the passage. Make a table: sentence, citation, supports / partly / doesn't / not in the folder, quote.Papers not in the folder yet?
Take three citations from the introduction, download the cited papers into literature/, then check whether each supports the sentence it is cited for. Quote the passage.- Run it on your own draft before you submit.
- “Not in the folder” is not “doesn’t exist”. Get those through the library, put them in literature/, and ask again.
A polished bibliography is not evidence of a correct bibliography.
11 Words you’ve heard
Skills, subagents, worktrees: words you’ve heard and half-wondered about. Each is a sentence and one thing to try. Unsure about a word that isn’t here? Ask the tool:
What is [a word you've heard]? Two sentences, for a professor. Then: would it help with my research?11.1 Skills
A routine, written down once: a name, when to use it, and the steps. The agent follows it when the task matches, or when you type its name. Both tools use the same format, and you never have to touch it:
Turn what we just did into a skill called [name].Then “What skills do you have?”, and next time, in another folder, just ask for the job. The referee panel (section 6) is a good first skill. More in section 13.
11.2 Connectors (MCP)
Your email, calendar and Drive, if you switch them on: section 17.
11.3 Subagents
The agent hands parts of a job to helpers that work in parallel, each with a fresh context; only their summaries come back. Faster, and it uses your limit faster.
Use subagents: one checks every number in the tables against the code, one checks the references, one compares the paper with the preregistration. Then give me one summary.11.4 The project file
AGENTS.md (and Claude’s CLAUDE.md): what the agent reads every time it starts in this folder. Section 13.
11.5 Plan mode
The agent reads, asks and proposes a plan, and changes nothing until you agree. For big or risky jobs. In the Claude app: choose Plan in the mode control under the prompt box. Codex has one too: ask it how to switch.
11.6 Context and /compact
Everything the agent knows in a session sits in one window (section 9). A long session fills it; /compact replaces the conversation with a summary, and /clear empties it. It compacts by itself when full, and details can go: for a long job, keep notes in a file. Or start a new session: they’re cheap.
11.7 Worktrees
A second folder for the same git project, on its own branch, so two agents can work at once without touching each other’s files. In the Claude app, turn on Worktree before you start a session. Or just ask:
Try this in a worktree; leave my folder as it is.11.8 GitHub and pull requests
A copy of the folder in the cloud, and a way to propose changes to a co-author: section 16.
11.9 Agents in the cloud
Choose Cloud instead of Local under the prompt box (or go to claude.ai/code; for Codex, chatgpt.com/codex), and the agent runs on a copy of your GitHub repository, on their computers. You can close the laptop and get a pull request back. The folder has to be on GitHub first, and the rules from section 7 apply: it runs on their machines.
12 Leave a trail
Turn a messy project into something a co-author could rerun, and check that they could. Practice on a messy one first, then do the same with a copy of one of your own folders.
The floor: the rerun test passes. In a new session, the agent reruns your folder from scratch and gets the same numbers.
12.1 The folder
A paper you’re writing with Kari Nordmann (fictional): Monthly or yearly? Price framing and willingness to pay. A survey experiment, cleaned by hand in Excel, analysed twice and written up in Word. In a session in an empty folder:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/messy-project.zip into this folder and unzip it.What you get:
messy-project/
├── survey_export_2026-03-14.csv the raw survey, 240 rows
├── data_clean_FINAL.csv cleaned by hand in Excel
├── analysis.py first try
├── analysis_v2.py the analysis
├── table1.txt results
├── table1 (copy).txt results?
├── notes.txt "the crazy WTP ones?? ask Kari"
└── draft_v3.docx the Word draft
Could Kari rerun this? Could you, next spring?
12.2 Organise
First a save point:
Set up git here and commit everything.Then:
Organise this folder so a co-author could rerun everything. Show me the plan first. Don't move anything yet.Argue with the plan: folder names, where the raw data goes, what happens to old versions. Nothing should be deleted; old versions can get a folder of their own. Then “Go ahead.”
Write a README for a co-author, and a project file for you. Then commit.The README is for people: what this is, and how to rerun it. The project file (AGENTS.md) is for the agent, which reads it every time it starts here.
12.3 The test
Start a new session in the same folder. It now knows nothing about this folder, like a new co-author.
Rerun everything from scratch. Do you get the same numbers?Not the same, or it can’t? That’s the useful part:
What's missing? Fix it, so the next rerun works.Using the practice folder? Compare the rerun with the numbers in the Word draft. They don’t match, and finding out why is the point of the exercise.
12.4 Why text
| Invisible | Visible | |
|---|---|---|
| Writing | Word: you can’t see what changed | Text: every change shows |
| Data | Clicking in Excel: no record | A script: the steps are the record |
Writing in LaTeX? That’s text already. Writing in Word? Keep Word for the paper, keep the rest in text, and ask for a Word copy when you need one: “Make a Word copy of README.md.”
You may remember Reinhart and Rogoff’s “Growth in a Time of Debt” (2010): above 90% debt to GDP, average growth of −0.1%. In 2013 a graduate student, Thomas Herndon, couldn’t reproduce it, asked for the spreadsheet, and got +2.2%. Part of the gap was an AVERAGE over the wrong rows, which left out five countries; the rest was years left out and the weighting. Your own rerun finds that kind of thing first. A wrong script you can read beats a right click you can’t remember.
13 Notes and structure
Where things go, and how the agent keeps track of a project for you.
13.1 Four questions
Answer these about a real project, in your own words:
- What do you re-explain every single time? → that goes in the project file.
- Where does undone work live? (If the honest answer is “my head” or “my inbox”, that’s the finding.) → a to-do file.
- Where are you allowed to be wrong? → notes, drafts, scratch.
- What could you never rebuild? → raw data, the manuscript. Protect it.
Everyone already has a filing system. It just isn’t written down, so it can’t be delegated to a co-author or an agent.
13.2 An example: my paper folder
Every paper of mine starts like this (a skill sets it up):
my-paper/
├── AGENTS.md the rules, for the agent
├── README.md for people: what this is, how to rerun it
├── STATUS.md where we are, what's next
├── manuscript/ everything that ends up in the PDF
├── code/ everything that computes
├── results/ the numbers the paper relies on
├── dev/ notes, drafts, talks: in git, not published
└── local/ downloaded papers, raw runs: not in git
My four answers: the project file (AGENTS.md), STATUS.md, dev/, and manuscript/ with results/. Yours will look different, and that’s the point of the next step.
dev/ or local/? Both hold things that aren’t the paper. The difference is git:
| In git and on GitHub | Published with the paper | |
|---|---|---|
manuscript/, code/, results/ |
yes | yes |
dev/ |
yes | no |
local/ |
no | no |
dev/is work in progress: notes, drafts, talks. It’s committed, so it’s backed up on GitHub and your co-authors see it, but it never goes out with the paper.local/is listed in.gitignore(section 8). Git never sees it, so it never reaches GitHub: downloaded PDFs you may not redistribute, big raw files, anything else that stays on your machine. To set it up: “Keep local/ out of git.”
13.3 Ask for a structure, not a template
Here is how I work: [your four answers]. Propose a folder structure for this paper and a project file that says where things go. Show me first; don't move anything.Argue with it. Then “Go ahead”, and “Commit” if you use git. Check nothing got lost: “How many files were there before, and how many now?”
13.4 Notes the agent keeps for you
At the end of each session, add what we did and what's next to notes/log.md. Put this rule in the project file.Keep literature/notes.md: one paragraph per paper we read, with the pages that matter.Next time you sit down:
Read the project file and the log, then tell me where we left off.13.5 The project file
A plain-text file in the folder that the agent reads at every start: what the project is, where things go, and your rules. Your profile is about you; the project file is about this folder.
One file: AGENTS.md. Codex and Claude both read it. One catch: if the folder also has a CLAUDE.md, Claude reads that instead, so keep just the one file. Both tools also read project files in the folders above the one you open, so open the project folder itself.
Getting one, and changing it. You don’t write it yourself; you ask:
Write an AGENTS.md for this folder: what the project is, where things go, and the rules you should follow. Show me first.Changes go the same way: “Add to AGENTS.md: never edit anything in data/raw.” Then read what it wrote: the first draft describes the folder, but only you know the rules.
What goes in:
- what this is, in one paragraph: the paper, the question, the data;
- where things go: raw data in data/raw, papers in literature/, scripts in code/, the paper is in paper/;
- the rules: never edit anything in data/raw; download papers into literature/, and give me the DOI when you can’t; only cite references with a verified DOI;
- how to rerun everything, in one command.
Keep out what it can see for itself (the file list), long essays, passwords, and anything about you (that’s your profile).
The habit that pays off. Corrected it twice about the same thing? Then:
Add that to the project file.Every correction you write down is one you never make again. Keep it short: it’s read at every start. It’s plain text, so open it in Notepad and fix it yourself whenever it’s wrong.
13.6 Skills, if you’re curious
Memory is the AI remembering you. A skill is you writing it down.
A skill is a saved workflow with a name and a note on when to use it. It’s worth making one for a routine you repeat, want done the same way every time, or want to hand to a colleague. You never touch the file format; you ask:
Turn what we just did into a skill called tidy-paper.From then on, “tidy this paper” runs the whole routine. Most things work fine without skills. Only bother for the ones you catch yourself repeating.
14 Your own work
Bring the real thing. Commit first (“Set up git here and commit everything”), then let it try.
A power check, no data needed:
Simulate a 2 × 2 experiment: the treatment effect is d = 0.3 for one product and d = 0.1 for the other. How many participants per cell do I need for 80% power on the interaction, not just the simple effects? Show the code and run it.Then ask what that means for the last study you ran.
A project folder: “What’s here? What’s missing? What would you do first?”
A dataset: “Describe this data. What looks odd?”
A draft: “Where is the argument weakest?”
A referee report you received: “Draft a response plan, point by point.”
A preregistration you’re writing: “What would a sceptical referee say I left open? Where could I still fork?”
Anything you’ve been putting off because it’s tedious.
Before you leave: what will you try on Monday? Pick one thing, and put it in your calendar. And don’t overthink it: just use it.
14.1 Worth reading
- Epoch AI benchmarks: where to track what the models can do.
- One Useful Thing, Ethan Mollick: the best running account of using these tools for academic work.
- The Plain Person’s Guide to Plain Text Social Science, Kieran Healy. Written for social scientists, not programmers.
- Good enough practices in scientific computing, Wilson et al. (2017). For researchers who aren’t software engineers.
- Spreadsheet horror stories, EuSpRIG. Real incidents with price tags.
Questions afterwards: jonas.moss@bi.no
15 Interviews
Eight interviews with founders who changed direction early, a codebook, and every quote checked. The transcripts are fictional and pseudonymised (F01 to F08), with an interview guide of six questions. The same steps work for consumer interviews or focus groups.
The floor: a codebook you argued with, and one theme whose quotes a script has checked word for word.
Your own transcripts work the same way, if your data plan allows a model to read them: what the agent reads is sent (section 7). Work on pseudonymised copies, in a folder of their own.
15.1 Get it, and a save point
In a session in your practice folder:
Download https://github.com/JonasMoss/vibing/raw/main/courses/2026-12-marketing/props/pivot-interviews.zip into this folder and unzip it. Then go into pivot-interviews, set up git, and commit everything.(Or copy practice/pivot-interviews from the course files.)
15.2 Is it really anonymous?
Read the transcripts. Is anything still identifying, even without the names? List it, with the file and line.Pseudonymising by hand misses things: a first name in passing, a firm you can recognise from its description. Here, three things slipped through.
15.3 A codebook first
Read the interview guide and three transcripts. Propose a codebook: six to eight codes, each with a definition and an example. Don't code anything yet.Argue with it. Merge two codes, split one, add the one you care about. Then “Save it as coding/codebook.md and commit.”
15.4 Code all eight
Code all eight transcripts with the codebook. Save every coded excerpt in coding/excerpts.csv: founder, line number, code, and the exact quote.What are the three main themes? For each: which founders, and the two best quotes. Save it as themes.md.15.5 Every quote, checked
Check every quote in themes.md and excerpts.csv against the transcripts, word for word, with a script. List any that don't match exactly.A script that compares text can’t be charmed by a confident paraphrase. This is the qualitative version of “rerun the code”: cheap, mechanical, and it lets you trust the rest.
15.6 Then push
- “Which founder doesn’t fit the themes? What does that do to theme 1?”
- “Does anyone contradict themselves?”
- “Did they plan the change, or does it only look planned afterwards?”
- “What would a sceptical quantitative reviewer say about theme 2?”
Done early?
Make a table: founder by code, with counts. Which codes always come together?Who does what. The agent reads eight interviews in a minute, proposes codes, finds every instance and checks every quote. You decide what a code means, judge the themes, and own the interpretation and the ethics.
16 GitHub, if you want it
git is the save points on your laptop (section 8). GitHub is a copy of that folder, with all its save points, in the cloud. It survives a dead laptop, lets a co-author work on the same folder, and records every change with a date and a name. OneDrive keeps files in sync; GitHub keeps every version.
Why a researcher might want it: backup with history; co-authors on one folder instead of emailed attachments; replication packages (journals increasingly ask for code and data, and a GitHub repository archived on Zenodo gets a DOI); course pages built from plain text.
16.1 Set up, once
Make a free account at https://github.com. Private repositories are free.
Ask the agent:
Install the GitHub command-line tool and log me in.It installs
ghand opens your browser to log in.
16.2 Put a folder on GitHub
Put this folder on GitHub as a private repository called [name]. Check .gitignore first: no PDFs, no raw data, nothing confidential.Then, whenever you’ve committed something worth keeping:
Push my latest changes.Laptop dead? On the new one, ask: “Download my repository [name] into this folder.” Everything is back, with its whole history.
16.4 Pull requests
Instead of pushing straight into your project, a co-author can propose a change: a pull request. You read it, discuss it, and accept it (merge) or not. The discussion and the revisions stay attached to the change. It’s referee-and-revise, for a folder.
Their side:
Make a new branch, [the change], commit it, push it, and open a pull request to [owner] that explains what you changed and why.Your side:
Are there open pull requests on this repository? Summarise them.Check out the pull request and test it. Does everything still run? Do any numbers change?Then “Comment on the pull request: [your question]”, or “Merge it, then pull the latest changes.”
16.5 Keep an eye on it
- GitHub tells you. Click Watch on the repository’s page, and you get an email (or a push notification in the GitHub mobile app) for every new pull request.
- A reviewer on GitHub itself. Connect the Claude or Codex app to the repository once. Then comment
@claudeor@codex reviewon a pull request, and it posts a review there.
16.6 Public or private
- Private by default. Only you and the people you invite can see it.
- Public only for what you’d publish anyway: a replication package, a course page.
- History is forever. A file you delete later is still in the history, so never commit passwords, respondents’ data or confidential data in the first place.
- Research data: check the data agreement before it goes to any cloud service.
17 Connectors: email, calendar, files
MCP is a standard plug that lets the agent use other services: your email, calendar or files, the same way it uses your folder. Each plug is called a connector.
Switch one on (Claude). Go to claude.ai/customize/connectors, pick Gmail, Google Calendar or Google Drive, click Connect, and approve with your Google account. Start a new session and type /mcp: it’s in the list. Codex supports MCP too; ask it how to connect yours.
Email:
Find all my emails from [name] this month. One line each: date, subject, what they want.Draft a reply. Show it to me before you send it.Calendar:
What does my week look like? Find two free hours for writing, and put them in my calendar.Drive:
Find the referee report on [paper] in my Google Drive, and draft a response plan, point by point.