# Lab 0.2: Your first hour, one control at a time (/how-ai-works/your-first-hour/lab) --- type: Document title: "Lab 0.2: Your first hour, one control at a time" description: "Chapter 0.2's lab: make the two privacy choices, then try each control on Brightline's practice files, from a file to a temporary chat." status: stable order: 12.7 ksor: owner: team:panaversity audience: [ public ] approval: by: process:panaversity at: 2026-10-29T06:28:52Z chapter: "0.2" part: "0" expert_status: null # as the Part 0 overview: Part 0 is optional on every route, after the skip quiz concepts: [] last_verified: 2026-10-17 sources: - id: anthropic-privacy resource: https://privacy.claude.com/en/articles/12109829-how-do-i-change-my-model-improvement-privacy-settings title: "How do I change my model improvement privacy settings? (Anthropic Privacy Center, updated 2026-08-03, verified 2026-10-16)" - id: anthropic-memory resource: https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context title: "Use Claude's chat search and memory to build on previous context (Claude Help Center, updated 2026-10-14, verified 2026-10-16)" - id: anthropic-incognito resource: https://support.claude.com/en/articles/12260368-use-incognito-chats title: "Use incognito chats (Claude Help Center, updated 2026-09-30, verified 2026-10-16)" - id: openai-data-controls resource: https://help.openai.com/en/articles/7730893-data-controls-in-chatgpt title: "Data controls in ChatGPT (OpenAI Help Center, updated 2026-09-28, verified 2026-10-16)" - id: openai-memory resource: https://help.openai.com/en/articles/8590148-memory-in-chatgpt title: "Memory in ChatGPT (OpenAI Help Center, updated 2026-09-19, verified 2026-10-16)" - id: openai-temporary resource: https://help.openai.com/en/articles/8914046-temporary-chat-in-chatgpt title: "Temporary chat in ChatGPT (OpenAI Help Center, updated 2026-09-19, verified 2026-10-16)" - id: anthropic-artifacts resource: https://support.claude.com/en/articles/17153992-what-are-artifacts-and-how-do-i-use-them title: "What are artifacts and how do I use them? (Claude Help Center, updated 2026-10-08, verified 2026-10-16)" - id: openai-library resource: https://help.openai.com/en/articles/20001052-using-library-to-manage-files-in-chatgpt title: "Using Library to manage files in ChatGPT (OpenAI Help Center, updated 2026-10-03, verified 2026-10-16)" - id: openai-uploads resource: https://help.openai.com/en/articles/8555545-uploading-files-and-audio-to-chatgpt title: "Uploading files and audio to ChatGPT (OpenAI Help Center, updated 2026-10-08, verified 2026-10-16)" generated: at: 2026-10-29T06:28:52Z by: process:panaversity trust_tier: unverified build_id: sha256:e1fbcb7d9a8f8b9b7cb96875022325c02c9ba4be94cef13b301cbd39a9edad85 dirty: true ksor_version: 0.0.60 --- In this lab you try the chapter's controls on Brightline's practice files, one control at a time. It takes about 30 minutes, on your free Claude or ChatGPT account. No account yet? Open a free one first, as 0.2.1 shows: claude.ai for Claude, chatgpt.com for ChatGPT. **What you will learn.** To make the two privacy choices, then use each control for the job it does: a file, a document, code, a stronger model, a project, memory and a temporary chat. And to tell where a result lands: in the chat, in a document you keep, or in another app. ## Before you start Download [`brightline-lab-ch0-2.zip`](https://github.com/panaversity/agentfactory-v2-resources/releases/latest/download/brightline-lab-ch0-2.zip), and unzip it. Its `files` folder holds three practice files: - `ap-policy-v3.pdf`: Brightline's policy for accounts payable, or AP, version 3, on two pages. - `invoices-this-week.csv`: Brightline's twelve unpaid invoices, as a spreadsheet file. - `vendor-email.txt`: an email from Maple Paper, one of Brightline's vendors. Brightline Wholesale Supply is the book's invented company, so the files hold nothing real. Use them, and put nothing from your own work into the chat. The files are dated September 2026, so the AI may say that the dates have passed. That is fine. If you use ChatGPT's free plan, do Lab 0.3 on another day. The free plan allows only a few chats with files in a few hours. Use your free accounts, in the browser or in the desktop apps from [0.2.4](desktop-apps.md). If you set the Claude app to use another company's model, go back to your Claude account first, as 0.2.4 shows. Attached files may not work in that mode. Copy this table onto paper or into a notes file. After each task, write in a few words what the control did for you. There is a filled-in table at the end, to compare with yours. | Task | What it did for you | | --- | --- | | 1. Two choices | | | 2. A file | | | 3. A document | | | 4. Code | | | 5. A stronger model | | | 6. A project | | | 7. Memory and a temporary chat | | | 8. Where it landed | | ## Task 1. Make the two choices Decide two things before your first message. To open Settings on either app, click your name at the bottom left. First, may your chats train the AI vendor's later models? In Claude, that is "Help improve our AI models", in Settings > Privacy.[^anthropic-privacy] In ChatGPT, it is "Improve the model for everyone", in Settings > Data controls.[^openai-data-controls] If you are not sure, turn it off. Second, is memory on? In Claude, it is "Generate memory from chats", in Settings > Memory.[^anthropic-memory] In ChatGPT, it is "Enable memory", in Settings > Personalization.[^openai-memory] Write both choices in row 1. [0.2.1](where-it-goes.md) shows both screens. ## Task 2. Answer from a file Open a new chat. Attach `ap-policy-v3.pdf` with the + button, and send this: ```text Does a $7,800 invoice need the controller's approval? Quote the line you used. ``` Then open the PDF, and find that line yourself. **What you should see.** Yes, with line 4.1 quoted: "Invoices over $5,000.00 need the controller's approval before payment." The file was in the window, so the model could read it. The line you found in the PDF is the check, because a guess from training can look the same ([0.2.2](the-controls.md)). ## Task 3. Keep a document Stay in the same chat, because the policy is already in it. Send this: ```text Turn this policy into a one-page checklist for the AP team, as a document I can edit. ``` Then find the document again, as if you came back tomorrow. **What you should see.** On Claude, the document opens beside the chat, and you can edit it there. Claude keeps it under Artifacts, in the sidebar.[^anthropic-artifacts] On ChatGPT, it may come as a Word file to download, which you open in Word or Google Docs. ChatGPT keeps the files it makes in Library, in the sidebar[^openai-library] ([0.2.2](the-controls.md)). ## Task 4. Let it run code Open a new chat for this new job. Attach `invoices-this-week.csv`, and send this: ```text Which of these invoices have a due date on or before Friday, September 18, 2026? Add up their amounts by running code, and show me the code. ``` **What you should see.** Six invoices, $14,439.10 in all. One of them, from Midwest Packaging, was already due on September 11, and the code still counts it. On ChatGPT, the code is folded away. Click Analyzed, above the answer, or a small `` icon after it. You may see "Analysis errored" first, when the AI found a mistake in its own code and fixed it. Then check the list of invoices it gives you. Every due date should be on or before September 18 ([0.2.2](the-controls.md)). ## Task 5. A stronger model Stay in the same chat. On ChatGPT, click Think in the message box. On Claude's free plan, Sonnet, the stronger of its two models, is already chosen. Open the model menu next to the send button to see it. Then send this: ```text County Freight offers 2% off invoice 5214, $8,400.00, if we pay it within 10 days instead of 30. We would borrow the money for those 20 days at 12% a year. Is paying early still worth it, and by how much? ``` **What you should see.** Yes, by about $113.87: the $168.00 discount, less about $54.13 of interest. With Think on, ChatGPT may take longer before it answers. Use the stronger choice when the question is hard or the answer matters ([0.2.2](the-controls.md)). ## Task 6. Make a project Make a practice project named "Brightline AP", and give it the policy and one instruction. On Claude, open Projects in the sidebar, then New project. Add `ap-policy-v3.pdf` under Context, and the instruction under Instructions. On ChatGPT, click the + beside Projects in the sidebar. Add the PDF under Sources, with Add sources, then Library: task 2 already put it there, so it costs no upload. Add the instruction in Project settings, from the ... button, and click Save. The instruction: ```text In every answer, name the policy line you used. ``` Then open a new chat inside the project. Open `vendor-email.txt`, which opens in a text editor such as TextEdit or Notepad. Copy all of it, paste it into the chat, and send it with this question under it: ```text What should we do with this email? ``` **What you should see.** Do not change the bank details because of the email. Call Maple Paper back on the phone number already in the vendor record, as line 5.1 says. The chat used the policy and the instruction, though you attached nothing in it. A project loads its files and instructions into every chat inside it ([0.2.2](the-controls.md)). ## Task 7. Memory, and a temporary chat Open your memory settings, as in task 1, and read what the app has saved about you. On a new account it may still be empty: Claude's shows "No files yet". Then open a temporary chat. On Claude, click the ghost icon at the top right of a new chat.[^anthropic-incognito] On ChatGPT, click the dashed circle at the top right.[^openai-temporary] Send this, then close the chat, and look at your chat list: ```text Is this chat kept in my chat list? ``` **What you should see.** Memory holds notes from your chats, or nothing yet. The temporary chat is not in your chat list, whatever the AI answered. The model may not know what kind of chat it is in, so your chat list is the check. A temporary chat stays out of your history and out of training, and the AI vendor still keeps a copy for up to 30 days ([0.2.1](where-it-goes.md)). ## Task 8. Sort what you made A result can land as an answer in the chat, as a document or file you keep, or as a change in another app, such as an email sent or a record updated. Write in row 8 where each of these three landed: - the answer about the $7,800 invoice, from task 2 - the checklist, from task 3 - the answer about Maple Paper's email, from task 6 Then add one more line to row 8: did anything in this lab change another app? **What you should see.** The $7,800 answer is in the chat. The checklist is a document you keep: under Artifacts on Claude, or in Library on ChatGPT. The Maple Paper answer is also in the chat. Nothing changed in another app. No email went to Maple Paper, and no bank details changed anywhere: the AI said what to do, and nobody did it. A reply it drafted for you is still only text in the chat. Your settings, your project and any memory note changed only inside the AI app. To change another app, an AI needs a connector that can act there, and this lab used none. When one does, check the change in that app itself. ## What your table may look like Your words will be different. | Task | What it did for you | | --- | --- | | 1. Two choices | Training off or on, and memory off or on, both written down | | 2. A file | It answered from the policy, and I found line 4.1 in the PDF | | 3. A document | A checklist I can edit, kept under Artifacts or in Library | | 4. Code | Six invoices, $14,439.10, worked out by code I could see | | 5. A stronger model | About $113.87, worked out step by step | | 6. A project | It used the policy without my attaching it, and named line 5.1 | | 7. Memory and a temporary chat | My memory notes, or none yet, and no row for the temporary chat | | 8. Where it landed | Two answers in the chat, the checklist kept as a document, and no change in any other app | If you have time, do the tasks on the other AI vendor too, with [0.2.3](both-ai-vendors.md) as your map. ## If something goes wrong - **You cannot find a control.** Search the settings for its name. The names and places change often, and 0.2.3 shows where each one is today. - **ChatGPT says you are out of uploads.** Its free plan allows three file uploads a day,[^openai-uploads] and this lab uses two: in tasks 2 and 4. Wait until the limit resets, and go on from your task. - **ChatGPT pauses the chat.** Its free plan limits chats that include files, and it may pause them for a few hours. Wait until the time it shows, or go on with Claude. - **It says it cannot run code.** Ask it to add the amounts step by step, and check its total against $14,439.10. - **You reach a usage limit.** Free plans limit how much you can send. Wait until it resets, and go on from your task. [^anthropic-privacy]: How do I change my model improvement privacy settings?, Anthropic Privacy Center. [^openai-data-controls]: Data controls in ChatGPT, OpenAI Help Center. [^anthropic-memory]: Use Claude's chat search and memory to build on previous context, Claude Help Center. [^openai-memory]: Memory in ChatGPT, OpenAI Help Center. [^anthropic-artifacts]: What are artifacts and how do I use them?, Claude Help Center. [^openai-library]: Using Library to manage files in ChatGPT, OpenAI Help Center. [^anthropic-incognito]: Use incognito chats, Claude Help Center. [^openai-temporary]: Temporary chat in ChatGPT, OpenAI Help Center. [^openai-uploads]: Uploading files and audio to ChatGPT, OpenAI Help Center.