Getting set up

Getting set up is signing in to Claude, picking between the browser, the desktop app and the phone, and changing two settings before you start. It takes four minutes and most of this lesson is the two settings, because both of them are easier to get right now than to discover in three weeks.

Sign in

Go to claude.ai and make an account. The free tier covers everything part 1 asks you to do: projects, file uploads, artifacts, memory, incognito chats and the code-execution setting are all on it. Where a paid plan buys you more of something, the lesson says so where it comes up.

There are desktop and mobile apps as well. Which platforms and where to download them is on the site, linked from the account menu, and it moves around often enough that a course should not be telling you. They are the same account and the same conversations, synced.

Which one should you use?

They are the same product, so this is about where you happen to be:

The browser for anything involving a document you are also looking at, anything you will copy the result out of, and anything with an artifact in it. The most screen, the least friction.

The desktop app if you use it daily. The only real advantage is that it is a window rather than a tab, which sounds trivial and is the difference between a tool you open and a tab you lose.

The phone for reading, for a question that occurs to you away from a desk, and for photographing something and asking what it says. It is bad for anything you intend to edit.

Nothing in this course needs a particular one.

Two settings worth changing now

Instructions for Claude. Click your initials, then Settings, then Instructions for Claude. These apply to every conversation on your account, on every plan. Whatever you would otherwise type at the top of every chat goes here once.

Good things to put in it, and the test is that each one is something you would otherwise repeat:

• What you do, in one line. “I run stock control for a small warehouse.”
• How you want answers shaped. “Answer in British English. Be concise. If I ask for a document, give me the document and not a description of it.”
• What to avoid. “Do not start replies by restating my question.”

Keep it short. It is read before every conversation, and a page of preferences dilutes the two that matter.

Memory. Go to Settings > Memory now, before you have used it for anything. The panel lists every entry Claude has saved about you, grouped by category, and you can delete any single one or turn the whole feature off from there.

Look at whether it is currently on, because that varies and this course is not going to guess for you. Then decide. On is more useful and means Claude accumulates a picture of you. Off is more predictable and means every conversation starts from the same place. If you leave it on, the thing to do is read that panel occasionally rather than never.

The ghost

Start a new chat outside a project and there is a ghost icon in the upper right. That is incognito, it is on every plan, and a chat started that way is kept out of your history and out of memory in both directions: it does not add entries and it does not read the ones you have. You get a black border and an Incognito chat label so you know which kind you are in.

Two things it is not. It is not deletion: an incognito chat is still retained for thirty days by default, or longer if your organisation has set a longer period. And it is not reversible; you cannot turn one into a normal chat afterwards.

So use it for the conversation you would rather not have sitting in a list on a shared screen, and for when you want a clean slate with no accumulated context leaning on the answer. It is worth knowing this exists before you need it rather than after.

What running out looks like

Every plan has a cap on how much you can send in a rolling window, and the free tier reaches it soonest. The exact numbers move, so the only thing worth writing down is the shape: it resets on a rolling basis rather than at midnight, and it counts what you send rather than how many chats you open.

The practical consequence, and the only reason it is in this lesson: long conversations cost more than short ones, because the whole conversation is re-sent every time you send a message. Somebody who starts a new chat per task gets more done in a day than somebody running one enormous chat, on the same plan. Lesson 4 is why.

Your turn

Three things, in order, and all three are quick.

1. Write your Instructions for Claude. Three lines maximum: what you do, how you want answers shaped, one thing to avoid.

2. Open a new chat and ask something where the instruction would show:

   Write me two sentences about why stock takes go wrong.

3. Open your memory panel and read whatever is in it.

Check: the answer in step 2 obeys the instruction you wrote in step 1. If you asked for British English, it spells it that way. If you asked for concision, you got two sentences and not five. Now delete one line from your instructions, ask the identical question in a new chat, and watch that part of the behaviour go away. That is the whole mechanism, and you have just seen both ends of it.

Recap

Same account everywhere. Browser for anything you will edit or copy out of, desktop if you use it daily, phone for reading and for photographing things.

Two settings on day one: standing instructions for whatever you would otherwise type at the top of every chat, and a look at the memory panel so you know it exists. Incognito for the conversation you do not want in the list.

Long conversations cost more than short ones, on every plan, because the whole thing is re-sent each time.

Next: the difference between a request that works and one that does not, which is four things and none of them are magic words.